fix prepare temp tables fails if production table has no values

[KatunaNorbert]

Fixes #1038.

• Fixed lake raw drop and lake etl drop so that they both work with pdr lake raw drop ppss.yaml sapphire-network 2023-01-01
• You can now drop rows from raw, or etl, and then resume building from where you left off, getting all the data loaded from the CSV
• You can very quickly execute different loops of:

1. update raw and view results, then drop part of raw: raw update -> describe -> raw drop partial
2. update raw again and verify it went to full: raw update (to full) -> describe (see full)
3. now update etl (to build full bronze tables) -> describe (see full bronze tables)
4. now run lake raw update to get latest from gql and sync raw data to duckdb
5. then run lake etl update to update the bronze tables based on the latest data you just fetched... only processing the new rows that have been fetched.
6. or you can lake etl drop everything, then run lake etl update and have your bronze tables rebuild in a second.

• ETL is using the right etl_tables (start_ts) and raw_tables (end_ts) mixture to calculate how to process data, and only process the new data required. Both bronze SQL queries also enforce this.

[idiom-bytes]

Very happy to see this test written @KatunaNorbert!!! This covers the use-case bang on, and gives me confidence that things are working like we expected them to.

Note: Please look at the fixtures inside test_etl.py. Perhaps try re-writing this test after to use the existing setup and reduce the amount of code you need to write.

[idiom-bytes]

Fantastic work @KatunaNorbert

Thank you for isolating everything in here, writing test coverage, and writing enough details throughout the ticket/PR for me to review and complete everything with ease