Align moment magnitude overload defaults with the implementations

[lispandfound]

Fixes #90

Problem

The @typing.overload stubs put the default on the Literal[False] variant while both implementations default bold_m to True:

# before
@typing.overload
def moment_to_magnitude(moment: float, bold_m: typing.Literal[True]) -> BoldM: ...
@typing.overload
def moment_to_magnitude(moment: float, bold_m: typing.Literal[False] = False) -> Mw: ...
def moment_to_magnitude(moment: float, bold_m: bool = True) -> BoldM | Mw:

So a bare moment_to_magnitude(m) resolved to the Literal[False] overload and was typed Mw, while the value returned at runtime is a BoldM. magnitude_to_moment had the same mismatch.

Mw and BoldM are conventions this module deliberately keeps apart in the type system, and the difference is not notional:

moment_to_magnitude(1e19)        = 6.6334   <- actual (BoldM)
moment_to_magnitude(1e19, False) = 6.6000   <- what the stub claimed
                      difference = 0.0334 magnitude units, ~5% in moment

Verification with ty (the checker in the dev group)

reveal_type on the three call forms, plus a BoldM passed positionally:

expression	before	after
moment_to_magnitude(1e19)	Mw ❌	BoldM ✅
moment_to_magnitude(1e19, True)	BoldM	BoldM
moment_to_magnitude(1e19, False)	Mw	Mw
magnitude_to_moment(BoldM(6.6))	error: Expected Mw, found BoldM	clean

Changes

Runtime behaviour is untouched — this is a stub-only change.

• Move = True onto the Literal[True] overloads.
• Drop the default from the Literal[False] overloads, so passing False stays explicit (a default on both variants would make a bare call ambiguous).

Verification

• pytest tests/test_moment.py — 14 passed.
• ruff check, ruff format, numpydoc lint clean.

🤖 Generated with Claude Code

[lispandfound]

Pushed a follow-up commit: the corrected overload defaults made ci / typecheck fail, and the failure was a real bug the stubs had been hiding.

tests/test_moment.py:202 constructs an Mw but omitted bold_m:

moment_newton_metre = moment.magnitude_to_moment(Mw(5.0))

The implementation defaults bold_m=True, so the BoldM formula was applied to an Mw value. Under the old stubs a bare call resolved to the Mw, Literal[False] overload, so the checker saw nothing wrong; with the defaults corrected it reports Expected BoldM, found Mw — exactly the mismatch that was there all along.

The test only asserts that a later KeyError is raised, so the wrong magnitude never affected the result and nothing caught it.

Fixed by passing the convention explicitly: magnitude_to_moment(Mw(5.0), bold_m=False).

ty check over the whole project now reports the same 13 diagnostics as main (all pre-existing; my local ty is newer than CI's, which is why main is green on CI). This is a small piece of evidence that the change does what it is supposed to: it makes the checker able to see convention mix-ups at call sites.