We should investigate if a shard split should be aborted if there's a concurrent reconfig.

Once the correct behavior is agreed upon and implemented, a test for the various corner cases, based on `tenant_migration_concurrent_reconfig.js` should be created.

N.B. It seems the desired behaviour would simply be to ensure the shard split fails cleanly (goes into aborted state, no data corruption/etc) when a reconfig occurs at various steps of the shard split.