Update submodules

[github-actions[bot]]

It's time to update our forks and submodules. Check for upstream updates and merge them if necessary.

Don't forget to use guide.

Subbmodules to update:

• [x] inference/ggml
• [x] inference/llama.cpp/llama.cpp
• [x] inference/whisper/whisper.cpp

[iboB]

perf sampling was improved in llama.cpp upstream which break our build. Working on it

[iboB]

found crash in ggml (bug). Fix, fork, redirect, make PR

https://github.com/ggerganov/ggml/pull/973

[iboB]

llama: disabling context shift may be relevant https://github.com/alpaca-core/llama.cpp/commit/441b72b91f818fe69497e5816f87969e90c73c43 but we already cover this

Still this begs the question remains whether we should handle a full context with an exception or allow non-exceptional break path.

We will decide if it becomes an issue.