Honor disable_numparse while wrapping column widths

[ShamikOfficial]

Fixes #448

While wrapping cells, _type(cell, numparse) passed numparse into has_invisible. disable_numparse was therefore ignored, and a value such as 100,120 was parsed as a number.

The flag is now passed as numparse=numparse.

Tested with the new regression test and the existing disable_numparse tests.

[killtheprompt]

The keyword argument fixes the disable_numparse=True case. There is still a default-options case from #448 at d3abd94:

from tabulate import tabulate

tabulate([["100,120"]], tablefmt="grid")  # succeeds
tabulate([["100,120"]], tablefmt="grid", maxcolwidths=20)
# ValueError: invalid literal for int() with base 10: '100,120'

Adding disable_numparse=True to the second call now succeeds. Local parametrized checks confirm that distinction; the existing suite also passes (383 passed, 1 skipped, Python 3.13).

Related: #449 already includes a fallback for the conversion failure and tests both settings. If this PR is intentionally limited to honoring disable_numparse, it may help to clarify that scope rather than mark all of #448 as fixed.

Prepared with Codex assistance; the checks above were executed locally against this PR's head.