Tokenize endpoint empty array crash fix

[przepeck]

🛠 Summary

CVS-186166
Adding checks for empty nested array, which was making OVMS crash

🧪 Checklist

• Unit tests added.
• The documentation updated.
• Change follows security best practices.
  ``

[Copilot]

Pull request overview

Fixes a crash in the tokenize request parser when the text field contains an empty nested array (e.g. {"text":[[]]}), by explicitly rejecting empty inner arrays and adding regression tests across tokenize endpoint test suites.

Changes:

• Add an explicit guard in TokenizeParser::parseInput() to reject empty inner arrays before indexing array[0].
• Add negative HTTP tests to ensure text with empty nested arrays returns an error (and does not crash) for embeddings, rerank, and LLM tokenize endpoints.
• Remove a previously existing GTEST_SKIP() in the LLM tokenize “add_special_tokens” test for VLM models.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 5 comments.

File	Description
src/tokenize/tokenize_parser.cpp	Adds empty-inner-array validation to prevent out-of-bounds access when inspecting nested arrays.
src/test/reranknode_test.cpp	Adds HTTP regression tests covering empty nested arrays for the rerank tokenize endpoint.
src/test/llm/tokenize_endpoint_test.cpp	Adds HTTP regression tests for empty nested arrays across all LLM tokenize test parameters; also removes a VLM-specific skip for add_special_tokens.
src/test/embeddingsnode_test.cpp	Adds HTTP regression tests covering empty nested arrays for the embeddings tokenize endpoint.

[dkalinowski]

Is it working for all paths, or only special token path? Do other tests have it unskipped?