[SPARK-46079][INFRA] Install torch nightly only at Python 3.12 in Infra docker image

### What changes were proposed in this pull request?

This PR aims to install `torch` nightly distribution and `torcheval` for Apache Spark 4.0.0 preparation.

### Why are the changes needed?

Currently, `pytorch` is not available in Python 3.12 environment.
- pytorch/pytorch#110436

We can start to use `nightly` build and switch to the official one via SPARK-46078 when it's released.

### Does this PR introduce _any_ user-facing change?

No. This is an infra only update.

### How was this patch tested?

1. Passed the image building.

![Screenshot 2023-11-23 at 4 39 42 PM](https://github.com/apache/spark/assets/9700541/5347ab49-15de-4bf1-a429-a5ecacd16d3a)

2. Verified manually
```
$ docker run -it --rm ghcr.io/dongjoon-hyun/apache-spark-ci-image:master-6975332773 python3.12 -m pip freeze | grep torch
torch==2.2.0.dev20231123+cpu
torcheval==0.0.7
```

### Was this patch authored or co-authored using generative AI tooling?

No.

Closes #43988 from dongjoon-hyun/SPARK-46079.

Authored-by: Dongjoon Hyun <[email protected]>
Signed-off-by: Dongjoon Hyun <[email protected]>

Author: dongjoon-hyun

dev/infra/Dockerfile

@@ -136,3 +136,6 @@ RUN apt-get update && apt-get install -y \
 RUN curl -sS https://bootstrap.pypa.io/get-pip.py | python3.12
 RUN python3.12 -m pip install numpy 'pyarrow>=14.0.0' 'six==1.16.0' 'pandas<=2.1.3' scipy unittest-xml-reporting plotly>=4.8 'mlflow>=2.8.1' coverage matplotlib openpyxl 'scikit-learn>=1.3.2'
 RUN python3.12 -m pip install 'grpcio==1.59.3' 'grpcio-status==1.59.3' 'protobuf==4.25.1' 'googleapis-common-protos==1.56.4'
+# TODO(SPARK-46078) Use official one instead of nightly build when it's ready
+RUN python3.12 -m pip install --pre torch --index-url https://download.pytorch.org/whl/nightly/cpu
+RUN python3.12 -m pip install torcheval