Don't cache-from Docker Hub twice

When we're building the image to test it, we can attempt to use the
upstream Docker Hub image, but when we're ready to publish it, we can
rely on the GHA docker cache (which will be warm from the test step)
removing a chunk of requests and (hopefully) reducing rate-limiting
errors (100 requests per 6 hours).

Author: frison

.github/actions/build-test-publish/action.yml

@@ -138,6 +138,8 @@ runs:
           prefix=${{ inputs.docker-label-prefix }},type=sha,enable=true,priority=100,format=short
 
     # https://github.com/docker/build-push-action
+    # Note this only uses GHA caches as they should we warm from the previous step.
+    # This helps with 429 rate limiting errors from Docker Hub.
     -
       name: 🚀 ${{ inputs.display-as }} (${{ inputs.platforms }})
       if: ${{ inputs.publish == 'true' }}
@@ -154,7 +156,6 @@ runs:
           type=gha,mode=max
         cache-from: |
           type=gha,mode=max
-          type=registry,ref=${{ inputs.docker-image }}:${{ inputs.docker-label-prefix }}latest,mode=max
 
     -
       name: 🖋 Docker Hub