feat(pymongo): Set MongoDB tags directly on span data

[0Calories]

Another PR to iron out some inconsistencies between the Python and Node SDKs. OTel attaches MongoDB tags directly on the span data, rather than in the tags databag. This change is necessary so that Relay can parse MongoDB spans in Python for metrics extraction.

Node SDK Spans:

Relevant MongoDB data is set directly as span data

Python SDK Spans:

Relevant MongoDB data is set in the tags databag

[codecov]

Codecov Report

Attention: Patch coverage is 66.66667% with 1 line in your changes missing coverage. Please review.

Project coverage is 75.98%. Comparing base (301c4b8) to head (9a2c4a3).
Report is 12 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3290      +/-   ##
==========================================
- Coverage   79.40%   75.98%   -3.42%     
==========================================
  Files         132      133       +1     
  Lines       14281    15218     +937     
  Branches     2999     3291     +292     
==========================================
+ Hits        11340    11564     +224     
- Misses       2094     2798     +704     
- Partials      847      856       +9     

Files	Coverage Δ
sentry_sdk/integrations/pymongo.py	60.34% <66.66%> (-23.69%)	⬇️

... and 9 files with indirect coverage changes

[szokeasaurusrex]

@0Calories couldn't this change be breaking in some way? What if a user has filters/dashboards/etc in Sentry set up based on the tags – these would break with your proposed changes.

[0Calories]

@szokeasaurusrex That's a good point, but consistency is still more important so that we can extract metrics for multiple SDKs. I believe the Python SDK in the near future will be implementing OTel's SDK under the hood, just like the JS SDK, so this exact change will occur and break dashboards for users who rely on these tags anyways

[szokeasaurusrex]

@0Calories I think this change will still require a major version bump if it is breaking. I understand that consistency matters, but semantic versioning dictates a major version bump for any breaking change.

We are in progress with implementing POTel in Python and intend to release a major when we do so, so I guess we can include this change then.

[szokeasaurusrex]

I suppose the change could be made backwards-compatible by setting this data both as tags and as data on the span. The data would of course be redundant, and we would probably need to consider implications in terms of storage costs on the server, but then we could include this change in a minor without violating semantic versioning

[0Calories]

That makes sense to me, I'll update this to include the data on both tags and data

[szokeasaurusrex]

Thanks for making the changes! Just a few more small things, and then I think the PR will be ready for merging

[0Calories]

Thanks for the review and suggestions @szokeasaurusrex, updated the PR with the assertions and your comment

[szokeasaurusrex]

Looks good!

[0Calories]

Awesome thanks @szokeasaurusrex! Could you also please merge the PR for me? It seems I don't have permission to do so on my own

[szokeasaurusrex]

@0Calories the PR needs to be updated with the base branch before you can merge