Update forks and submodules

[iboB]

We really need some resolution on #33 ... until then adding issues manually.

We should do this soon and again in the second half of October

[iboB]

Windows-only build error in ggml-cuda: Since it was small and was already done in llama, instead of finding a revision that works, I made a temporary branch of our fork and attached the submodule to it: https://github.com/alpaca-core/ggml/compare/master...alpaca-core:ggml:tmp-fix-cuda

[iboB]

llama_reset_timings is gone. Comment out our uses and fork to a different issue

[iboB]

whisper.cpp missing newly added argument to ggml_graph_plan. nullptr seems safe. Instead of redirecting all submodules, make a temporary branch to our fork and attach the submodule to it:

https://github.com/alpaca-core/whisper.cpp/compare/master...alpaca-core:whisper.cpp:tmp-ggml_graph_plan

[iboB]

sampling logic in llama is completely changed. Our sampler doesn't compile. Working on this

[iboB]

15 can be resolved as min_keep is documented and optional

[iboB]

Applying changes for:

• https://github.com/alpaca-core/llama.cpp/commit/faf69d4237c9ae4d7f572b4674d1002463e8acd3
• https://github.com/alpaca-core/llama.cpp/commit/4af8420afba39de4968adf2695ce12fd42422a13