in https://jira.mongodb.org/browse/SERVER-96195 we discovered a bug where bulkWrite with a multi:true update vs an update command with a multi: true update increment statistics differently. both of the code paths eventually call underlying functions in write_ops but have different paths before getting there. we should merge these paths and namespaces to reduce the chance where a bug is fixed in one but not the other.

since bulkWrite is essentially a superset of the insert, update, and delete commands, we should be able to make the parsing logic a little more generic and have, for example, the update command get parsed into a list of BulkWriteCRUDOps and call bulk_write::performWrites() directly. hopefully without adding too much overhead