fix(python): move accelerate and GPU-specific libs to build-type (#3194)

Some of the dependencies in `requirements.txt`, even if generic, pulls
down the line CUDA libraries.

This changes moves mostly all GPU-specific libs to the build-type, and
tries a safer approach. In `requirements.txt` now are listed only
"first-level" dependencies, for instance, grpc, but libs-dependencies
are moved down to the respective build-type `requirements.txt` to avoid
any mixin.

This should fix #2737 and #1592.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>

backend/python/exllama2/requirements-cpu.txt

@@ -0,0 +1,3 @@
+transformers
+accelerate
+torch

backend/python/exllama2/requirements-cublas11.txt

@@ -1,2 +1,4 @@
 --extra-index-url https://download.pytorch.org/whl/cu118
 torch
+transformers
+accelerate

backend/python/exllama2/requirements-cublas12.txt

@@ -1 +1,3 @@
 torch
+transformers
+accelerate

backend/python/exllama2/requirements.txt

@@ -1,4 +1,3 @@
-accelerate
 grpcio==1.65.4
 protobuf
 certifi