Add clear error message for missing SentencePiece model in get_spm_processor (fix #41553)

[AvinashDwivedi]

Fix: clearer error when SentencePiece model file is missing for Llama tokenizer

Fix:
Improved error handling in get_spm_processor to provide a clear, actionable message when the required SentencePiece model file is missing.

Before:
The tokenizer raised an unhelpful low-level

TypeError: not a string

from the SentencePiece library when vocab_file was None.

After:
It now raises a concise, human-readable error:

ValueError: Missing SentencePiece model file: 'None'. 
This tokenizer requires a .model file provided by the 'mistral-common' package. 
Install it with: pip install mistral-common

Fixes: #41553

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: llama

[AvinashDwivedi]

@ArthurZucker would you please help me why I'm seeing checks are failing although the changes are very small. I'm new to opensource So, please help me sir!

[Rocketknight1]

I think it's strange to reference mistral-common in the Llama tokenizer, but up to you @ArthurZucker @itazap!

[vasqu]

Copy pasting a bit but let's coordinate #41553 (comment)

[AvinashDwivedi]

@ArthurZucker @Rocketknight1 as I can see this issue is still not solved can you guys give me some idea how you want it? I'll make it exactly same as you want it.