fix(migrations): migration serialization bugs for PK fields

[seladb]

Description

Fixes 2 bugs that were discovered in #2176

Motivation and Context

Two bugs in migration serialization when a PK field is configured with
non-default attributes:

1. generated=False on PK fields was lost during serialization —
   Field.deconstruct() only emitted generated when the value was
   truthy. If a PK field explicitly set generated=False, the kwarg was
   omitted, causing the field to revert to its per-engine default
   (True for IntField) on deserialization. The fix includes the
   generated kwarg whenever self.pk is true, preserving the explicit
   value.
2. PK fields with source_field="<FIELD_NAME>" were dropped from CreateModel —
   MigrationWriter._format_create_model() collected all source_field
   values and skipped any field whose name appeared in that set. When a
   PK field had source_field="<FIELD_NAME>", its own name matched the set entry
   and it was incorrectly excluded from the migration output. The fix
   exempts PK fields from this deduplication logic.

How Has This Been Tested?

• Added tests for deconstructing with different options of generated in PK and non-PK fields
• Added test for a migration with source_field that is similar to the field name

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have added the changelog accordingly.
• I have read the CONTRIBUTING document.
• I have added tests to cover my changes.
• All new and existing tests passed.

[codspeed-hq]

Merging this PR will not alter performance

✅ 24 untouched benchmarks

---

_{Comparing seladb:bugfix-2176 (1e5401a) with develop (c0de49f)}

[seladb]

@waketzheng @abondar I'm not sure why the coverage workflow didn't start, can you cancel and restart it (I don't have permissions to do so)?