Issues and requirements for next test-tooling updates

[apostasie]

As we rewrite tests, missing pieces and bugs in the tooling will appear.
This is to collect them:

• mode private propagation is problematic, as each subtest is in its own isolated namespace - ideally, the subtest would by default be in the parent namespace
• async commands are clunky to use right now - implemented foreground and background, but it would be better to just have a setting on the command that would somehow switch from RunCmd to Start+WaitCmd - design TBD
• testutil.RequiresBuild and many others are missing: review them all
• having to use test.Case at the root of the test is not elegant (especially since it requires a redundant/useless description. Maybe ‘Setup’ could return a prepolulated test.Case instead
• use of data.Identifier in subtests hoping to get the parent testid is error prone - rethink Identifier() and maybe use data.set(mainId) or something in the line
• test.Group being parallel with no way to change it is confusing/error prone. Questioning the value of test.Group entirely.
• nerdtest.Command is annoyingly convoluted - rethink the use of Target and maybe remove it along with custom Run/Clone
• in line with above, some functions cannot be hidden with t.Helper - rethink that
• identifier hashing (sha1) method is unsatisfying - should go back to something readable
• missing something like .AddCleanup() when a routine inside Setup does want to add additional cleaning
• requirements are now called twice, because of the issue with Private

Test suites to rewrite:

• started: builder
• completion
• TODO: compose
• started: container
• in progress: image
• TODO: ipfs
• in progress: login
• network
• system
• volume
• main