@roji Pushed some updates. Updating data is still failing for TPH. Optimistic concurrency test 1 is still failing in all cases.

Found two more bugs while investigating this:

1. When updating, if a column hasn't changed (e.g. the discriminator), we don't create a parameter for it, and everything gets mismatched. This was raised by Andriy in Support sproc input/output parameters on non-concurrency-token properties efcore#28908 (comment), and I'm working on it there. It basically means that partial updates are currently broken.
2. We don't properly detect concurrency exceptions when the rows affected is in the result set, and there are other result set values to propagate.
   a. In regular SQL, when there's nothing to propagate, we expect a single value with 1 or 0 (from OUTPUT 1 SQL).
   b. But when there's something to propagate, we interpret the existence of a row as a sign that the update succeeded.
   c. While this works for regular SQL, with sprocs we always get back a row. So we need to add logic to check for 1 or 0 even when there's propagation.
   This was actually #28997.

I'll be back to full work on Tuesday (will be partially doing stuff on Monday too) and will concentrate on fixing all this (not feeling great about the work I've done here...)

@ajcvickers just making sure that everything works for you and you're not blocked on me by anything.