fix: parse_float(bool/int) and parse_int(float-string) return wrong results

[gaoflow]

What

Three related bugs in parse_float / parse_int that cause silent wrong results:

call	before	after
get_float(True)	0.0	1.0
get_float(False)	0.0	0.0
get_int(True)	True (bool)	1 (int)
get_int(False)	False (bool)	0 (int)
get_int("3.5")	0 (default)	3

Root causes

parse_float with bool/int input:
is_float(True) is False (bool is not a float subclass), so _parse_with falls through to str(True) → "True" → float("True") raises ValueError → None → caller's default (0.0).

parse_int with bool input:
is_integer(True) is True (bool is an int subclass), so _parse_with short-circuits and returns True unchanged — a bool, not an int.

parse_int with a float-string:
int("3.5") raises ValueError with no fallback, so numeric strings with a fractional part always fall through to the caller's default.

Fix

# parse_float: short-circuit bool/int before string conversion
def parse_float(val: str) -> float | None:
    if isinstance(val, (bool, int)):
        return float(val)
    return _parse_with(val, type_util.is_float, _parse_float)

# parse_int: explicit bool cast + float-string fallback
def _parse_int(val: str) -> int | None:
    try:
        return int(val)
    except ValueError:
        try:
            return int(float(val))
        except ValueError:
            return None

def parse_int(val: int | str) -> int | None:
    if isinstance(val, bool):
        return int(val)
    return _parse_with(val, type_util.is_integer, _parse_int)

All 756 existing tests pass with the corrected expectations.