ref(grouping): Cache enhancements in split form

[lobsterkatie]

Right now, even when we're using split enhancements, the version we cache isn't split, which means that every time we pull them out of the cache, the work to split the rules has to be done all over again. This is obviously not ideal, so this PR changes the way we compute (and parse) the cached string so that it now includes the split rules.

Since the rust enhancer can only parse base64 strings including one set of rules, this is done by computing a separate string for each of the three kinds of rules - rules in their original/pre-split form, classifier rules, and contributes rules - and then concatenating them, separated by a character which can't ever appear in base64. Older base64 strings containing only the original rules and no delimiter can still be parsed, because for them, splitting on the delimiter will be a no-op.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

⚠️ Parser warning

The parser emitted a warning. Please review your JUnit XML file:

Warning while parsing testcase attributes: Limit of string is 1000 chars, for name, we got 2083 at 1:157235 in /home/runner/work/sentry/sentry/.artifacts/pytest.junit.xml

Additional details and impacted files

@@             Coverage Diff             @@
##           master   #92000       +/-   ##
===========================================
+ Coverage   36.78%   87.63%   +50.84%     
===========================================
  Files        9662    10359      +697     
  Lines      540758   587563    +46805     
  Branches    22604    22604               
===========================================
+ Hits       198930   514885   +315955     
+ Misses     341407    72257   -269150     
  Partials      421      421               

[yuvmen]

looks good 👍