Identifying the table/schema on the backend when ingesting from a generic mutation handler.

I have an app with a generic mutation handler as per https://github.com/TanStack/db/blob/main/docs/index.md#2-electricsql-sync

When I POST the mutations to the server, the only way of identifying the table/schema they should be applied to is to look in the syncMetadata.relation. This is current set from the ?table=... param in the streamOptions for the ElectricCollection.

This is fragile because it's totally valid not to use the table param when making a request to a proxy.

I see two ways forward:

1. make a more resilient way of assigning the tablename / server-side identifier to the collection and have it reliably passed through to the mutations
2. require the user to pass through the information, e.g.:

import genericMutationFn from './mutations'

const addTodo = useOptimisticMutation({ mutationFn: (params) => genericMutationFn('todos', params)})

As per #96, I would advocate for just having a way of passing the identifier through, e.g.: with an additional optional property on the collection options.

Start with docs/index.md#2-electricsql-sync, then trace ElectricCollection streamOptions, syncMetadata.relation, and useOptimisticMutation through the generic mutation handler. Review the related discussion in #96 before choosing the identifier API. Done means the backend can reliably identify the target table or schema without depending on the table query parameter, with the chosen behavior documented.