Partially revert "ci: Use mv instead of cp in upload step"

This partially reverts commit fe7c97c.

I kept a mv, not a cp, for the one that shuffles major artifacts around,
because the size of those artifacts are big enough to matter, sometimes.
I don't think the diagnostic info will be that heavy, by comparison.

src/ci/scripts/upload-artifacts.sh

@@ -23,14 +23,14 @@ if [[ "${DEPLOY-0}" -eq "1" ]] || [[ "${DEPLOY_ALT-0}" -eq "1" ]]; then
 fi
 
 # CPU usage statistics.
-mv build/cpu-usage.csv "${upload_dir}/cpu-${CI_JOB_NAME}.csv"
+cp build/cpu-usage.csv "${upload_dir}/cpu-${CI_JOB_NAME}.csv"
 
 # Build metrics generated by x.py.
-mv "${build_dir}/metrics.json" "${upload_dir}/metrics-${CI_JOB_NAME}.json"
+cp "${build_dir}/metrics.json" "${upload_dir}/metrics-${CI_JOB_NAME}.json"
 
 # Toolstate data.
 if [[ -n "${DEPLOY_TOOLSTATES_JSON+x}" ]]; then
-    mv /tmp/toolstate/toolstates.json "${upload_dir}/${DEPLOY_TOOLSTATES_JSON}"
+    cp /tmp/toolstate/toolstates.json "${upload_dir}/${DEPLOY_TOOLSTATES_JSON}"
 fi
 
 echo "Files that will be uploaded:"