[py-tx] Validate hash length

[aryzle]

Summary

fixes #1609

validates hash length in SignalExchangeAPIWithSimpleUpdates (used by stopNCII, file, etc)

Test Plan

reproduce error by adding "00000000000000000000000000000000" to the list of example PDQ hashes and try to run tx fetch

...
Building index for pdq with 17 signals...
Traceback (most recent call last):
  File "/home/vscode/.local/bin/tx", line 33, in <module>
    sys.exit(load_entry_point('threatexchange', 'console_scripts', 'tx')())
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspaces/ThreatExchange/python-threatexchange/threatexchange/cli/main.py", line 321, in main
    inner_main()
  File "/workspaces/ThreatExchange/python-threatexchange/threatexchange/cli/main.py", line 314, in inner_main
    execute_command(settings, namespace)
  File "/workspaces/ThreatExchange/python-threatexchange/threatexchange/cli/main.py", line 157, in execute_command
    command.execute(settings)
  File "/workspaces/ThreatExchange/python-threatexchange/threatexchange/cli/fetch_cmd.py", line 163, in execute
    DatasetCommand().execute_generate_indices(settings)
  File "/workspaces/ThreatExchange/python-threatexchange/threatexchange/cli/dataset_cmd.py", line 292, in execute_generate_indices
    index = index_cls.build(signals.items())
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspaces/ThreatExchange/python-threatexchange/threatexchange/signal_type/index.py", line 216, in build
    ret.add_all(entries)
  File "/workspaces/ThreatExchange/python-threatexchange/threatexchange/signal_type/pdq/pdq_index.py", line 75, in add_all
    self.index.add(
  File "/workspaces/ThreatExchange/python-threatexchange/threatexchange/signal_type/pdq/pdq_faiss_matcher.py", line 240, in add
    self.faiss_index.add_with_ids(numpy.array(vectors), numpy.array(i64_ids))
...

after rebuilding tx with the fix in SignalExchangeAPIWithSimpleUpdates, error goes away as the offending entry is ignored (note the count 17 went down to 16)

...
Building index for pdq with 16 signals...
Index for pdq ready
...

[Dcallies]

For test plan, can you confirm if you were able to generate a successful reproduction by adding an invalid length hash in the PDQ samples?

[aryzle]

@Dcallies not yet, I'm trying to get a unit test going, but I'll do a manual test too

[aryzle]

seems like some tests do catch this 😂 will dig into them later today

[Dcallies]

blocking: This string could be quite long, suggest truncating it and doing a "..."

[aryzle]

@Dcallies I commented out the bad entry here because a few tests depended on it (DatasetCommandTest and test_utils in test_vpdq_faiss.py because VPDQSignal.get_examples calls PdqSignal.get_examples) and didn't make sense to include a bad entry in get_examples when it's expected to have only valid examples. We can def refactor later to handle this, or make another method that returns mixed/dirty examples to test on.

[Dcallies]

Hmm, would like to have a regression test that would exercise this condition, but you've been at it for a while, let's merge this one.