Unable to download required artifacts (plc_wrapper_main and DREAMPlace tarball) from GCS (HTTP 403)

Hi!

I am trying to set up `circuit_training` and hit a blocker downloading required artifacts from the documented GCS paths.

**Date tested:** February 19, 2026 (UTC)
**Environment:** Docker on macOS arm64 host (linux container)

**Repro**
```
curl -sS -D - -o /dev/null \
https://storage.googleapis.com/rl-infra-public/circuit-training/placement_cost/plc_wrapper_main

curl -sS -D - -o /dev/null \
https://storage.googleapis.com/rl-infra-public/circuit-training/dreamplace/dreamplace_python3.9.tar.gz
```
**Result**
Both return **HTTP/2 403** with:
`AccessDenied: Anonymous caller does not have storage.objects.get ... (or it may not exist)`

I also confirmed `/usr/local/bin/plc_wrapper_main` becomes a 298-byte XML file when downloaded via Docker, causing:
`OSError: [Errno 8] Exec format error.`

**Request**
1. Could you clarify the supported path for external users?
2. Are these artifacts intentionally private now?
3. If access is required, what is the process to obtain it?

Thanks! Grace

[Artifact_Access_FAIL_logsnippet.md](https://github.com/user-attachments/files/25427565/Artifact_Access_FAIL_logsnippet.md)

Start by reproducing the two documented GCS URLs with the curl commands in the issue and inspect Artifact_Access_FAIL_logsnippet.md. Check how the Docker setup obtains /usr/local/bin/plc_wrapper_main and the DREAMPlace tarball. Done means the supported external-user access path is established and documented, or the artifact references are corrected and downloads no longer return HTTP 403.

Written by the indexing model from the issue text.