fix: handle random_state=0 correctly in leiden clustering

[Ekin-Kahraman]

Summary

• Fix falsy check on random_state in _cluster() that caused random_state=0 to skip setting the RNG seed
• Changed if random_state: to if random_state is not None: so that 0 (the default) correctly calls optimiser.set_rng_seed(0)
• Added regression tests verifying seed is set for random_state=0 and skipped for random_state=None

Root cause

In muon/_core/tools.py line 1005, the condition if random_state: evaluates to False when random_state=0 because 0 is falsy in Python. This meant the default value (random_state=0) and any explicit random_state=0 call never set the RNG seed, producing non-deterministic clustering results.

Closes #154

Test plan

• test_leiden_random_state_zero_sets_seed — verifies set_rng_seed(0) is called when random_state=0
• test_leiden_random_state_none_skips_seed — verifies set_rng_seed is not called when random_state=None

[Ekin-Kahraman]

CI status update:

I've pushed a fix for the test failures in tests/test_leiden_random_state.py — the tests were using unittest.mock.patch("leidenalg.Optimiser") which requires leidenalg to be installed, but CI doesn't have it. Now the tests mock leidenalg at the sys.modules level so they work without the package installed.

The remaining CI failures are pre-existing issues unrelated to this PR:

• Lint (Black): muon/_atac/preproc.py and muon/_atac/tools.py need reformatting under black==26.3.0. These files are not touched by this PR.
• Build (test_muon_preproc.py): 8 failures in TestInPlaceFiltering with TypeError: 'mappingproxy' object does not support item assignment at muon/_core/preproc.py:827. Also not touched by this PR.

Our change is a single-line fix in muon/_core/tools.py line 1005 (if random_state: → if random_state is not None:). Happy to rebase if needed.

[Ekin-Kahraman]

@gtca @ilia-kats friendly ping on this — CI has been green since the Apr 11 fix. Small change (+96/-1), straightforward bug (random_state=0 was being treated as falsy). Happy to rebase or close if it isn't on the roadmap.

[ilia-kats]

This is basically a subset of #166 and I would prefer to merge that one, but I'd really like some feedback from @gtca first.

[Ekin-Kahraman]

Happy to defer to #166. The regression test here (test_leiden_random_state_zero_sets_seed, mocked leidenalg call to assert set_rng_seed(0) fires) might be worth lifting into #166 since it doesn't currently have one. Happy to open it as a follow-up PR against #166's branch once it lands, or leave it here as a reference.