Add batch support to /anonymize and /deanonymize endpoints

[bijay-odyssey]

Closes #1045.

Problem

/analyze already accepts text as either a single string or an array of strings, returning a matching array of results when given a batch (via BatchAnalyzerEngine). /anonymize and /deanonymize never got the same treatment — even though the core package already ships BatchAnonymizerEngine and BatchDeanonymizeEngine, unused by the HTTP layer.

Change

Both endpoints now accept text as a string (unchanged, existing behavior) or an array of strings. In array mode:

• analyzer_results (for /anonymize) / anonymizer_results (for /deanonymize) must be an array of the same length as text, one list of per-text results.
• The response becomes a matching array of results (same shape as the single-text response, one object per input text).
• A length mismatch between text and the results array raises a clear 422 instead of silently misaligning or dropping data.
• Omitting the results field entirely in batch mode defaults each text to an empty results list (text returned unchanged), matching the existing single-text default-empty behavior.

I went with looping directly over the engines in app.py (rather than reusing BatchAnonymizerEngine.anonymize_list/BatchDeanonymizeEngine.deanonymize_list) because those methods return only the anonymized text string, dropping items — which would make the batch response shape inconsistent with the single-text response. Looping directly preserves full response parity between single and batch modes.

Testing

Added presidio-anonymizer/tests/test_app.py (no HTTP-level test file existed for this service before) using Flask's test client, covering:

• Single-text behavior is unchanged
• Batch anonymize returns one result per text
• Batch anonymize with missing analyzer_results defaults correctly
• Batch anonymize with mismatched lengths returns 422
• Batch deanonymize, including a full encrypt → decrypt round trip
• Batch deanonymize length-mismatch validation

Full existing suite: 282 passed, 0 regressions. ruff check clean.

Docs

Updated docs/api-docs/api-docs.yml (request/response schemas + a batch example) for both endpoints, mirroring exactly how /analyze's existing batch support is already documented there. Validated with openapi-spec-validator.