gh-104066: Improve performance of hasattr for module objects

[itamaro]

Fixes gh-104066

microbenchmark on main

python -m pyperf timeit -s 'import os' 'hasattr(os, "getenv")' ..................... Mean +- std dev: 37.8 ns +- 1.5 ns 

python -m pyperf timeit -s 'import os' 'hasattr(os, "nothing")' ..................... Mean +- std dev: 541 ns +- 16 ns 

microbenchmark with PR

python -m pyperf timeit -s 'import os' 'hasattr(os, "getenv")' ..................... Mean +- std dev: 37.2 ns +- 2.6 ns 

python -m pyperf timeit -s 'import os' 'hasattr(os, "nothing")' ..................... Mean +- std dev: 55.8 ns +- 3.0 ns 

the missing attr case is almost 10x faster

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @carljm for commit 8c07ffb 🤖

If you want to schedule another build, you need to add the 🔨 test-with-refleak-buildbots label again.

[carljm]

This looks fine to me; it's following the existing pattern already used to optimize hasattr for both types and normal objects, by avoiding raising and then suppressing an AttributeError.

Will leave it for a couple days before merging in case any other reviewers want to take a look.