Install no cuda stuff, cpu only for tf and torch

Setup on Windows and Ubuntu takes ages because we install gpu enabled versions of tensorflow and torch, while very few students have a gpu. Especially knowing that torch is only there for the Transformers day.

- Use tensorflow-cpu
- Specify wheel location for torch so it downloads a cpu versions
- Remove all unnecessary nvidia-cu... packages

This reduces the virtualenv size with 4 Gb!
```bash
9.4G	./lewagon
5.4G	./lwlight
```

specs/releases/linux.txt

@@ -127,18 +127,18 @@ nltk==3.7
 notebook==6.4.12
 notebook_shim==0.2.0
 numpy==1.23.4
-nvidia-cublas-cu12==12.1.3.1
-nvidia-cuda-cupti-cu12==12.1.105
-nvidia-cuda-nvrtc-cu12==12.1.105
-nvidia-cuda-runtime-cu12==12.1.105
-nvidia-cudnn-cu12==8.9.2.26
-nvidia-cufft-cu12==11.0.2.54
-nvidia-curand-cu12==10.3.2.106
-nvidia-cusolver-cu12==11.4.5.107
-nvidia-cusparse-cu12==12.1.0.106
-nvidia-nccl-cu12==2.18.1
-nvidia-nvjitlink-cu12==12.2.140
-nvidia-nvtx-cu12==12.1.105
+# nvidia-cublas-cu12==12.1.3.1
+# nvidia-cuda-cupti-cu12==12.1.105
+# nvidia-cuda-nvrtc-cu12==12.1.105
+# nvidia-cuda-runtime-cu12==12.1.105
+# nvidia-cudnn-cu12==8.9.2.26
+# nvidia-cufft-cu12==11.0.2.54
+# nvidia-curand-cu12==10.3.2.106
+# nvidia-cusolver-cu12==11.4.5.107
+# nvidia-cusparse-cu12==12.1.0.106
+# nvidia-nccl-cu12==2.18.1
+# nvidia-nvjitlink-cu12==12.2.140
+# nvidia-nvtx-cu12==12.1.105
 oauthlib==3.2.2
 opt-einsum==3.3.0
 packaging==21.3
@@ -223,7 +223,7 @@ tenacity==8.1.0
 tensorboard==2.10.1
 tensorboard-data-server==0.6.1
 tensorboard-plugin-wit==1.8.1
-tensorflow==2.10.0
+tensorflow-cpu==2.10.0
 tensorflow-datasets==4.6.0
 tensorflow-estimator==2.10.0
 tensorflow-io-gcs-filesystem==0.27.0
@@ -239,7 +239,8 @@ toml==0.10.2
 tomli==2.0.1
 tomlkit==0.11.5
 toolz==0.12.0
-torch==2.1.0
+# torch 2.1.0 cpu only on the next line
+https://download.pytorch.org/whl/cpu-cxx11-abi/torch-2.1.0%2Bcpu.cxx11.abi-cp310-cp310-linux_x86_64.whl
 tornado==6.2
 TPOT==0.11.7
 tqdm==4.64.1