fix(pcap): _make_magic ignores nanosecond with explicit endian flags; _read_protos ignores size

[JarryShaw]

Describe the bug
misc/pcap/header.py:343-349: when lilendian/bigendian are given, _make_magic returns the microsecond magic whatever nanosecond says. header.py:317 _read_protos(size) ignores size, and nothing calls it.

Reproduction

Arguments	Got	Expected
(lilendian=True, bigendian=False, nanosecond=True)	4d3cb2a1	nanosecond little-endian magic
(lilendian=False, bigendian=True, nanosecond=True)	a1b2c3d4	nanosecond big-endian magic
byteorder='big', nanosecond=True	a1b23c4d	a1b23c4d (correct)

_read_protos with size of 0, 4 or 99 always returns (4, {'lilendian': True}).

Expected behavior
The endian-flag branch honours nanosecond. _read_protos is fixed or removed.

System information
main at b10e7951b, CPython 3.14.7, Linux.

Additional context
Found during the #719 second-round sweep, then reproduced on main.

[JarryShaw]

Correction to the reproduction table above, measured by the #1112 worker. On main, the endian-flag branch of _make_magic always returns the nanosecond magic for little-endian and the microsecond magic for big-endian. So the little-endian failure is nanosecond=False returning 4d3cb2a1; nanosecond=True already returned the right value. The big-endian row was right as filed.