Update methodology so that we keep provider information in IP and OP data

[yiwen-h]

In order for inequalities methodology to work in PBM we need to keep provider in the IP and OP data, so we can join with the correct inequalities factors

We currently overwrite this (https://github.com/The-Strategy-Unit/nhp_model/blob/13c08ab3b398ca9579e54037cd9954f0f3f15ce5/src/nhp/model/data/databricks.py#L168-L169)

Fixes TODO at https://github.com/The-Strategy-Unit/nhp_model/blob/13c08ab3b398ca9579e54037cd9954f0f3f15ce5/src/nhp/model/activity_resampling.py#L154-L156

[tomjemmett]

the reason for doing this was to prevent issues occurring at the aggregation stage for IP, and at runtime for OP (collapse the rows).

we may struggle to get PBM to work at national level with inequalities due to compute constraints