Use copy CSV instead of insert for pandas to postgres

[bendnorman]

This PR:

• Improve performance by chaniging the method argument of to_sql calls to use a function from the pandas docs that uses COPY CSV instead of INSERT.
• Creates a helper function that enforces datatypes on dataframes based on the types specified in the metadata.
• Changes data_mart.fossil_infrastructure_projects.cost_millions column from an Int to Float because I was getting a "could not cast int to float" error when enforcing the types.

On my comp make etl_local went from 6m39s -> 2m58s.

[TrentonBush]

Does this work if the df columns are in a different order than the metadata definition? Just noticed the csv is written separately from the COPY statement that defines the headers

[bendnorman]

Good question! I tested out mixing up the order of tables, and df.to_sql() still inserted the data with the correct column order.