api: round-trip injects a spurious <identification>/<encoding> not in the source

[webern]

Surfaced by the reworked round-trip classifier (PR #225; part of #212).

What

When a source file has no <identification> (or no <encoding>), the api writer emits one anyway. add:encoding and add:identification are the two highest-yield blockers on the corpus.

Verified on ksuite/k007a_Notations_Dynamics.xml:

• expected: no <identification>
• actual: <identification><encoding /></identification> (an empty encoding)

lysuite/ly33c_Spanners_Slurs.xml differs from its source by only the injected empty <encoding /> (distance 1 — fixing this alone flips it to PASS).

Impact (reorder-free candidate files; 550 total)

• add:encoding: 535 files (sole blocker of 3)
• add:identification: 420 files

Stop emitting an empty <encoding/>, and do not synthesize an <identification> when the source had none.

Repro

make dump-api-roundtrip && make classify-api-roundtrip
# build/api/classified.json: signatures add:encoding / add:identification

References

• Part of api: expand corpus coverage of the api round-trip pass-list #208; feeds api: implement ranked fixes and grow the round-trip corpus baseline #213
• Per api: define divergence policy and guardrails for round-trip corpus coverage #214, decide whether absence should round-trip as absence
• Relates to api reorders identification/encoding children on round-trip #220 (encoding-child ordering) and Add Encoding Supports Hints to ScoreData #98 (encoding supports hints)
• Surfaced by PR audit: classify api round-trip failures by measured divergence and rank a worklist #225