Update convert_hf_to_gguf: support Qwen3.5 embedding models

[SamMalayek]

Overview

Adds convert_hf_to_gguf.py support for Qwen3.5 embedding models using Hugging Face Qwen3_5TextModel, such as Rebine/Qwen3.5-Embedding-0.8B.

Additional Information

Manual testing:

• Real HF -> GGUF conversion
• Metadata: 24 blocks, LAST pooling, no MTP
• llama-embedding runtime load
• 1024-d embedding output
• Unsupported pooling fails closed

Did not add new test file to /tests (agents guidelines)

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: YES - AI was used to assist with investigation, implementation, and testing. I manually reviewed the changes and validated them.

[SamMalayek]

No changes made in recent force-push. Rebase only (also tested locally again):

https://github.com/ggml-org/llama.cpp/blob/master/CONTRIBUTING.md#after-submitting-your-pr

If your PR becomes stale, rebase it on top of latest master to get maintainers attention

[SamMalayek]

@CISC I rebased and the diff shows the patch is unchanged. I ran all the tests as well. Could you review this when you have time, or suggest another reviewer? Thanks.

[CISC]

You want to try that again?

[SamMalayek]

You want to try that again?

Mostly moved some code into a couple functions. No change in behavior.

Thanks again for the review.

[SamMalayek]

@CISC I’ve addressed the suggestions in the latest update. Ready for another review.