materialize-sql: validations and migrations for only-null columns

[williamhbaker]

Description:

WIP - The necessary plumbing is in place based on the struct updates from the updated flow pin and all existing tests pass. This can't be properly tested without a control plane build that populates write_inference though.

Handling for fields that have only null or no types, in both validate and apply.

Columns for these fields can be created based on a present write_inference. Schema inference may have only ever observed a null value, or have never observed a value at all, but we can still make a good guess on what kind of column to create based on the type from the write schema.

Similarly, we can always migrate these columns, since we know that they have only ever had null values. A possible scenario is this: The write schema has type: string, but the source field is almost always null (or not present). Eventually a value shows up, but it is inferred as a type: string, format: date-time. Although a TEXT column could not typically be migrated to a DATETIME column, we know all the existing values must be null so the column can be trivially migrating by dropping & re-creating it with the new type.

Workflow steps:

(How does one use this feature, and how has it changed)

Documentation links affected:

(list any documentation links that you created, or existing ones that you've identified as needing updates, along with a brief description)

Notes for reviewers:

(anything that might help someone review this PR)

---

This change is