fix(search): fallback to searching for errors if the bulk search across multiple datasets fails

[barkbarkimashark]

The search code fans-out and searches multiple datasets if the search criteria calls for it. But someones we encounter errors in the search when targetting the issue platform dataset and it errors the entire issue search. Instead, we should attempt the search, and fallback to only searching the errors dataset when we encounter an error.

[codecov]

Codecov Report

Merging #48310 (7af2f46) into master (18f867f) will increase coverage by 5.36%.
The diff coverage is 84.78%.

❗ Current head 7af2f46 differs from pull request most recent head 00164b8. Consider uploading reports for the commit 00164b8 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #48310      +/-   ##
==========================================
+ Coverage   75.51%   80.87%   +5.36%     
==========================================
  Files        4771     4772       +1     
  Lines      201807   201850      +43     
  Branches    11524    11524              
==========================================
+ Hits       152385   163255   +10870     
+ Misses      49167    38340   -10827     
  Partials      255      255              

Impacted Files	Coverage Δ
src/sentry/conf/server.py	93.75% <ø> (ø)
src/sentry/search/snuba/executors.py	85.30% <50.00%> (-9.72%)	⬇️
src/sentry/tasks/auto_ongoing_issues.py	95.65% <91.30%> (+52.31%)	⬆️
src/sentry/features/__init__.py	100.00% <100.00%> (ø)
src/sentry/issues/ongoing.py	100.00% <100.00%> (ø)

... and 603 files with indirect coverage changes

[evanh]

Why does the issue platform query fail?

[wedamija]

Why does the issue platform query fail?

This is more of a safety net when we have incompatible queries. They can fail for various reasons, usually cases where we have a mismatch between schemas, but sometimes it's when a filter just doesn't apply to the issue platform but we didn't account for it, so we end up erroring. Ideally we see these problems and fix them properly, but this allows us to avoid breaking the issue stream in the mean time.