Fix affected_rows data source for ON DUPLICATE KEY UPDATE operations

[ehasrouni]

Problem: affected_rows always returned 0 for ON DUPLICATE KEY UPDATE operations in Trilogy adapter.

Root Cause: In this PR, we had used raw_connection.affected_rows which returns stale/incorrect values for ON DUPLICATE KEY UPDATE in Trilogy.

The correct values are available in result.affected_rows from the execute result object, so this simply changes the implementation to make use of the result rather than the connection to fetch the affected_rows, super simple change.

This PR:
Switch from raw_connection.affected_rows to result.affected_rows when available
Preserve nil for adapters without support
Adds a test to cover for that edge case
Fix batch accumulation to handle nil values properly

[ehasrouni]

@jkowens hey! this should be the last piece of the puzzle (a small bug fix to what we merged last week). It just adds support for ON DUPLICATE KEY UPDATE operations that were returning 0 with the Trilogy adapter by simply using the result's affected_rows rather than the raw_connection. Business as usual for unsupported adapters.

Thanks a lot! let me know if you have any other concerns

[jkowens]

Thanks @ehasrouni 👍