gh-136602: Fix wrong offset_data being used for sparse files with an additional size PAX keyword

[mxmlnkn]

My attempt to fix #136602.

• Issue: Next TAR header offset recomputation is wrong for GNU sparse 1.0 file combined with 'size' PAX header key #136602

[python-cla-bot]

All commit authors signed the Contributor License Agreement.

[mxmlnkn]

_apply_pax_info suppresses all int-conversion errors:

cpython/Lib/tarfile.py

Lines 1628 to 1631 in 85ec3b3

	try:
	value = PAX_NUMBER_FIELDS[keyword](value)
	except ValueError:
	value = 0

Is this also necessary for this PR for int(pax_headers["size"])? It seems like it would break more than it helps, or are there use cases for this?