BFV noise bound ever-so-slightly underestimating on one test case

[j2kun]

See https://github.com/google/heir/actions/runs/14040657498/job/39309761604

==================== Test output for //tests/Examples/lattigo/bfv/noise/mult_indep_16_debug:multindep16debug_test
...
lattigo.bgv.mul
  [1]
  Noise: 36.28 Total: 125
  Noise Bound: 36.27 Gap: -0.01
--- FAIL: TestBinops (1.11s)
panic: Noise Bound Exceeded [recovered]
	panic: Noise Bound Exceeded

[ZenithalHourlyRate]

I suspect this will fail now and then...it do is a probablisitic (assumed to be 99.99%) bound but it should not fail that frequently. The problem should be with the underlying noise model or my implementation, but as far as I can tell there are quite a few questionable parts in every noise analysis (also implementation choices) so I think making them work completely should still be an undergoing research topic.

[j2kun]

In that test, maybe we can avoid this by just removing one of the operations so it's 15 instead of 16?

[ZenithalHourlyRate]

No, the error happens at the first multiplication so no matter how many ct there are it is probable to error again, so I disabled all 8,16,32 test in the PR.

[j2kun]

What if we add a pass option for validation, that does something like count of operations that failed noise validation, and only err if that count exceeds a value given in the pass flag. This would only be used for testing purposes.

[github-actions]

Note

Re-commenting because this issue was closed with unresolved TODOs.

This issue has 1 outstanding TODOs:

• tests/Examples/lattigo/bfv/noise/mult_indep_16_debug/BUILD:31: re-enable as non-manual

This comment was autogenerated by todo-backlinks