[Backsearch P1] Implement the matching algorithm

[lin-d-hop]

Description

After great research and investigations outlined in this spike, let's move forward implementing the matching algorithm to the level of 99.9% accuracy :raised_hands:

Acceptance Criteria

1. Implement the matching algorithm with appropriate error handling
2. Unit test coverage

[rogup]

The first step of this is done and we're wrapping up the work for now.

These changes https://github.com/DigitalCommons/property-boundaries-service/pull/15 are ready to be deployed to production.

And this further change is deployed on staging https://github.com/DigitalCommons/property-boundaries-service/commit/c2f70a068ea109990eca9919953989c1166a90ca

So the state of the pipeline is:

• On the night of 8th every month, a Borg backup of the PB database occurs.
• Once the Borg backup is done, it triggers the following 2 pipelines:

1. A pipeline on prod-2, which:
   • updates company ownership data, publishes to all LX users
   • downloads the latest raw INSPIRE data and backs it up to the Hetzner storage box
   • transforms the raw data and inserts it into the pending_inspire_polygons table, which can be viewed by LX super users only
2. A pipeline on staging-2, which:
   • updates company ownership data, publishes to all LX users on staging
   • downloads, transforms, and analyses the INSPIRE data and publishes those that successfully match for all LX users on staging. This is a simplified version of the pipeline that skips the segmentation/merge matching so make it faster. But it still takes ~24hrs and not sure how reliable it will be.