[model] add Command R7B support

[ojuanjo]

What does this PR do?

Fixes #6338

Adds support for CohereLabs/c4ai-command-r7b-12-2024.

• Registers CommandR-7B-Chat with Hugging Face and ModelScope download sources.
• Adds a dedicated cohere2 template without affecting the existing cohere template.
• Matches the official tokenizer format, including the BOS token, a required system turn, START_RESPONSE, and END_RESPONSE.
• Adds a standalone unit test and a gated-tokenizer alignment test.

Validation:

• Passed make style and make quality.
• Passed the license check.
• Passed the LLaMA Factory CPU test suite in tests/ and tests_v1/.
• Verified the cohere2 template against the model's official gated tokenizer.
• Completed local LLaMA Factory QLoRA SFT and DPO training on an RTX 5080; no training artifacts are included in this PR.

Before submitting

• Did you read the contributor guideline?
• Did you write any new necessary tests?