Support Custom Endpoints

Currently, bundle-cache hardcodes the endpoint that hosts the bucket [here](https://github.com/block/bundle-cache/blob/bcd00476d6d8e348723afbfb312435d95b824941/gradlecache/s3.go#L240) to the object store provided by AWS.

We have our own internal S3 cache and were curious if the project would be open to a contribution to support a custom endpoint (could be anything, GCP, AWS, etc), assuming what I mentioned above is correct. Are there any reasons why this wouldn't work (other places where AWS s3 assumptions are made, etc)?

Start with gradlecache/s3.go at the linked endpoint construction around line 240, then inspect nearby code for other AWS-specific assumptions. Done means defining how a caller supplies a custom object-store endpoint and confirming the cache can use an internal S3-compatible service without breaking the existing AWS path.

Written by the indexing model from the issue text.