GH-3289: Test conversion and writer configuration through CLI

[zhangxinyao88]

Rationale for this change

Continue #3289 with convert/to-avro coverage and observable writer configuration checks. Existing command tests largely assign fields directly and check file existence. Configuration-loading tests do not establish that a setting changes the resulting file.

What changes are included in this PR?

Add fourteen cases using deterministic Avro and Parquet inputs:

• Both conversion commands preserve record values and apply column projection, a supplied schema, and a selected codec.
• Existing output bytes are preserved without --overwrite and replaced by readable records with it.
• --property and properties/XML configuration files change a UUID field from string storage to 16-byte UUID storage in the Parquet footer.
• An explicit property overrides the corresponding file setting.

Assertions read records, codec metadata and physical/logical schema types. Paths contain spaces; no help-text or full-log snapshots are added. Read, inspection and CSV scopes in #3858/#3859/#3860 remain separate.

Are these changes tested?

./mvnw -o -B -ntp -pl parquet-cli spotless:check test: 97 cases, 95 passed, 2 skipped, no failures/errors. All fourteen new cases passed; Spotless and RAT checks passed. Java 17 and cached dependencies were used; no full-reactor or multi-JDK build was run.

Eleven temporary local regression controls triggered assertion failures when projection, supplied schemas, codecs, overwrite or configuration propagation/precedence was ignored. All production edits were restored before the successful module run.

Are there any user-facing changes?

No. Tests only.