Several runtime optimizations for CubicODSDelta

[MaxAlex]

Adds several performance optimizations to delta_ods.py:

• Increases MAX_MERGE_RECORDS to 500k, approximating the limit on fact_ods.py
• The 'more_pending' log flag is now based on whether we hit the merge records limit, rather than a separate scan of the fact tables
• When reading in CDC changes, instead of a single query for full data that uses LIMIT to restrict batch size, do two stages: first, a query just against the header__change_seq to find the upper and lower bound values that match the limit, and then a second query using those bounds. The previous version, which used ORDER BY header__change_seq LIMIT {limit} would require reading and sorting all header__change_seq values across all tables, rather than sorting only values that land within the interval.
• Restrict the merge predicate based on the odin month/year, so that the merge process is only touching files that need to be updated. This only works for updates which include the edw_inserted_dtm, but that looks like it's always the case (for tables that include that column.)

[ealexa05]

Confirming this as a safe change, will go ahead with dev performance test