Fix(snowflake): extract date from datetime values, quote audit's this_model

[georgesittas]

These changes ensure that we can successfully run the quickstart example in Snowflake (tested).

The two issues I faced when I tried to run sqlmesh plan for it were:

1. Got the error Failed to cast variant value .... to DATE when trying to evaluate the seed model. The problem was that the dataframe we pass in write_pandas had a column of type datetime64[ns], but the underlying column's type was DATE (ds's type). See related SO discussion which I included in a comment here.

2. After fixing the above issue, I got another error when trying to evaluate the full model, specifically when trying to run its assert_positive_order_ids audit. The reason was that in Audit.render_query we were creating a query with a table whose name is produced by the snapshot's table_name method. Since the resulting name is prefixed with the (case-sensitive) physical schema, we would then normalize it in the renderer's render method and hence produce an invalid schema reference. For example, sqlmesh__SQLMESH_EXAMPLE.<model_name> would be converted into SQLMESH__SQLMESH_EXAMPLE.<model_name>, which didn't exist. I fixed this error by enforcing quotes in said name before rendering, which I believe should be ok because the model name is already normalized, only the schema isn't.