ref(alerts): Catch more errors for anomaly detection

[ceorourke]

While doing prod testing for anomaly detection and expecting to hit Seer timeouts, nothing prevented the alerts from being created or edited despite the service not being available yet. I suspect there are errors other than the specific ones we're watching for, so we should check for what those could be and handle them appropriately.

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 78.24%. Comparing base (36ca8d1) to head (ac10df8). Report is 431 commits behind head on master.

Additional details and impacted files

@@             Coverage Diff             @@
##           master   #75617       +/-   ##
===========================================
+ Coverage   64.74%   78.24%   +13.50%     
===========================================
  Files        6825     6829        +4     
  Lines      303109   303351      +242     
  Branches    52178    52210       +32     
===========================================
+ Hits       196233   237358    +41125     
+ Misses     100261    59610    -40651     
+ Partials     6615     6383      -232     

Files	Coverage Δ
src/sentry/incidents/logic.py	90.98% <100.00%> (+38.31%)	:arrow_up:
src/sentry/seer/anomaly_detection/store_data.py	91.56% <100.00%> (+21.43%)	:arrow_up:

... and 1688 files with indirect coverage changes