Fix!: Always recreate materialized view

[VaggelisD]

Materialized views in engines like BigQuery and Snowflake are invalidated if the underlying table(s) are replaced. Toy repro scenario:

1. CREATE TABLE tbl AS (SELECT 1 AS col);

2. CREATE MATERIALIZED VIEW mview AS (SELECT * FROM tbl);

3. CREATE OR REPLACE TABLE tbl AS (SELECT 2 AS col);

4. SELECT * FROM mview;

This flow can be mirrored in the following toy SQLMesh project:

# Full model
MODEL (name example_schema.t1, kind FULL); SELECT 1 AS col;

# Materialized view
MODEL (
  name example_schema.t2,
  kind VIEW (
    materialized true
  )
);

SELECT * FROM example_schema.t1;

Running the very first plan will yield the following error in either engine, e.g:

google.api_core.exceptions.BadRequest: 400 Materialized view <t2> references table <t1> which was deleted and recreated. The view must be deleted and recreated as well

This is because, during the plan:

1. The physical table for t1 is created as a dry run
2. The materalized view for t2 is created as a dry run
3. The physical table for t1 is recreated/replaced as part of the FULL insertion strategy
4. The materialized view exists, and based on the pre-existing ViewStrategy it is not recreated

This PR alters (4) such that any materialized view with upstream dependencies is recreated from scratch in order to restore correctness.