Make rsync-antora-reference more robust

[wilkinsona]

Would it be possible to add some retry logic to rsync-antora-reference or otherwise make it more robust? Boot's docs have failed to deploy for the last several days with a broken pipe. A couple of examples:

2025-04-01T06:59:53.0057635Z updating: 3.5-SNAPSHOT/api/java/org/springframework/boot/autoconfigure/task/TaskExecutionProperties.Mode.html (deflated 83%)
2025-04-01T07:01:20.8269788Z updating: ziST9yCn (deflated 2%)
2025-04-01T07:06:25.7475415Z client_loop: send disconnect: Broken pipe
2025-04-01T07:06:25.7540788Z   adding: ziMtU62V
2025-04-01T07:06:25.7553536Z ##[error]Process completed with exit code 255.

2025-04-01T06:58:49.7066524Z updating: 3.5-SNAPSHOT/api/java/org/springframework/boot/autoconfigure/task/TaskExecutionProperties.Mode.html (deflated 83%)
2025-04-01T07:00:01.9580585Z updating: ziST9yCn (deflated 2%)
2025-04-01T07:05:09.5126653Z client_loop: send disconnect: Broken pipe
2025-04-01T07:05:09.5193522Z   adding: ziMtU62V
2025-04-01T07:05:09.5207773Z ##[error]Process completed with exit code 255.

[rwinch]

I think this was caused to to exceedingly large directories as a result of gh-36 I'm closing this as a duplicate, but we can reopen it if that doesn't solve the problem