CLN: Catch more specific exceptions in groupby

[jbrockmendel]

Working hypothesis is that this will help turn up the problem underlying #27902. Regardless catching more specific exceptions is worthwhile.

[jbrockmendel]

Thoughts here? Conflicts are going to start building up with local branches addressing catching-too-much

[WillAyd]

Looks good - more specific is certainly better with these

[jreback]

can you run groupby perf benchamarks. some of the code you changes was a sore point for perf recently.

[jreback]

rebase & run groupby asvs

[jbrockmendel]

Rebased. asvs look like noise, which makes sense.

asv continuous master HEAD -E virtualenv -f 1.1 -b groupby
[...]
       before           after         ratio
     [91e5b85a]       [0c8ae4a9]
     <master>         <winapply>
+         111±3μs          170±2μs     1.53  groupby.GroupByMethods.time_dtype_as_field('float', 'shift', 'transformation')
+      10.0±0.7ms       12.9±0.2ms     1.28  groupby.Categories.time_groupby_nosort
+         121±4μs          135±6μs     1.12  groupby.GroupByMethods.time_dtype_as_group('object', 'any', 'direct')
-         174±4μs          121±4μs     0.70  groupby.GroupByMethods.time_dtype_as_group('datetime', 'any', 'transformation')

e.g. re-running time_groupby_nosort comes back with NOT SIGNIFICANTLY CHANGED.

[TomAugspurger]

Thanks!