refactor: Internal doc restructure

[AndrewSisley]

RELEVANT ISSUE(S)

Resolves #448, #395, #390, #443, #427, #387, #484, #495

DESCRIPTION

Reworks the internal data structure used to house documents within the planner package. Solves a number of existing issues and allows for further code improvements. Automated tests should not have changed (minus a couple of tiny exceptions), and a couple have been added. More may be added as I start looking at the linked issues.

Key changes include:

• Documents in the planner package are now []interface{}s, field location is now deterministic instead of map driven (inherently prevents the alias type bugs that drove me to do this now instead of in v0.4).
• New mapper.Select struct and friends to replace parser.Select, this is currently constructed from parser.Select as it seems quicker to get it working that way first - the parser package could be removed later once the consuming code is more stable. This should contain all the info nodes will need, and allowed the removal of a lot of the complexity from the aggregate constuctors and hopefully type joins, and references to the ast stuff from the planner package.
• RenderNode is no longer required and has been deleted.
• The connor package has been pasted into our repo and has been adapted to work with the new doc structure.

Very strongly recommend reviewing commit by commit - they should be clean:

• TEMP - enable logging caller will be dropped before merge
• Alias internal doc object can be glossed over as it is overwritten by the restructure commit, but might be useful for context
• Copy-paste connor fork into repo is a clean copy paste from our connor fork, no code changes were made, although a handful of build/CI related files were omitted
• Restructure internal document data structure contains the bulk of the work and should be reviewed very carefully - it is a very large commit and pretty much all of it is 'manual' coding - please also keep an eye out for anything that looks half done

To do/confirm/discuss before merge:

• Connor location and future. I've managed to strip away most of the changes to this package, but there are some, including a reference to []core.Doc as []core.Doc is a struct and is not handled by a switch-case []interface{}. I dont think we should change all inner collections to []*core.Doc (which would solve this), but you guys might have other opinions. UPDATE: decision made, connor will live in defra.
• Most/all of connor's unit tests have been disabled for now, possibly depending on whether connor remains in the defra repo or not, we should decide to delete or preserve them. I am in favour of deletion, and if you guys want to keep them I would be in favour of doing that in a new issue. UPDATE: these will be deleted.
• A handful of our unit tests have been disabled (func signatures changed). I see these as more of a liability that should be deleted, but am open to discussion here. No integration tests have been disabled (if you spot any please flag it). UPDATE: decision made, tests deleted.
• I'll run the benchmarks locally and check to see if there any significant changes. Performance is not a focus here though, but it is a significant change and might have an impact. RESULT: No significant changes, results in another comment.
• If connor is staying in the defra repo, cherry-pick the connor copy-paste commit into develop before merging this PR. Ensures proper credit is preserved regardless of how much it may then change.
• Expand code-documentation for mapper-types (raised by Fred)
• Document mapper 'entry' funcs
• Revisit mapper 'entry' func names
• Dont forget to remove the temp logging commit...
• Post merge, review outline query spec and update if necessary
• Post merge, create issue to abstract away core.Doc from external planNode interface refactor: Internal doc restructure #471 (comment)

Stuff I will be adding as I wait for reviewers (will merge as soon as properly approved, regardless of how many of these are done):

• I'll go through any linked issues that will likely be solved by this and add tests. In some cases I might tweak the prod code if required and effort is minimal. EDIT: Issues added individually below - resolving this task early.
• I would like to use this excuse to test out the coz profiler, will post results if I get round to this. UPDATE: I didnt get round to this
• I would like to rework the join code. I think there can be significant improvements there making use of this stuff.
• I would like to go through the parser package and drop any now-dead code. Is probably a fair bit in there that is not needed now, but I dont see this as a mandatory task pre-merge - feel very free to disagree with me here and ask that I do it before merging.
• If connor is staying in the defra repo, delete the dead code and look at potential simplifications. UPDATE: This has been broken out of this PR and can be found here: refactor: Trim imported connor package  #530
• Boost integration test coverage of filters - is really low. UPDATE: new issue for this Integration test coverage of filter conditions is really poor #511 - will do outside of this PR, but possibly before this PR is merged. UPDATE2: tests in PR - test: Restructure and expand filter tests #512

Related issues have been solved (commit per issue):

• Multiple aliased joins on the same relationship do not correctly handle filters #390
• Missing tests/probable bug: deep group filter #443
• Aggregate: Allow filter defined within aggregate for child relationships #427
• Aggregate: Handle multiple input types #387
• bug: including the subtype in the top level filter argument won't produce a MultiNode #484
• Mutation create doesn't return the version information of the newly created document #495

Super happy to have a call to walk through this if anyone feels the need. Is probably going to be a bit of a pain to review.

HOW HAS THIS BEEN TESTED?

Int. tests.

CHECKLIST:

• I have commented the code, particularly in hard-to-understand areas.
• I have made corresponding changes to the repo-held documentation.
• I have made sure that the PR title adheres to the conventional commit style (subset of the ones we use can be found under: tools/configs/chglog/config.yml

ENVIRONMENT / OS THIS WAS TESTED ON?

Please specify which of the following was this tested on (remove or add your own):

• Debian Linux

[AndrewSisley]

This cost is like max 10-15 mins more in my opinion on the odd occasion the plan graph is changed significantly. If the plan graph is changed with a small change (example: removal of renderNode) then the maintenance is a couple of seconds that would be taken to do that small change (remove that node for example, just 2 lines per test). On top of that #521 will further lessen the pain.

I disagree, and I think we are going in circles. Will leave these two deleted unless really really pushed back on.

Need to merge