Handle zero-bytecode coverage stats

[kimeta]

What does this PR do?

This fixes how library coverage stats interpret missing JaCoCo counters.

Previously, the stats pipeline treated two different cases the same way and reported all library coverage metrics as N/A:

• libraries that have no executable bytecode to cover
• libraries that do have bytecode, but were compiled without line debug information

Changes:

• when JaCoCo shows there are zero executable methods, library coverage is now recorded as fully covered 0/0 (1.0)
• when only line coverage is unavailable, instruction and method coverage keep their real values instead of being collapsed to N/A
• validator and tests were updated to accept and verify the zero-total full-coverage shape