fix: remove null URNs from census data

[PsypherPunk]

Context

Minor follow-up to #1748 to tidy some inconsistencies seen in the output data.

AB#242989

Change proposed in this pull request

• explicitly drop any null entries from the URN data before setting the index
• simplify the joining of pupil and workforce data (join will default to using the index)

Guidance to review

.dropna(subset=["URN"])

The workforce data contains a null row for some of the earlier years. Previously, this was stripped when the inner join didn't find a match with the pupil data. However, following #1688, the outer join keeps this in place resulting in an index that is a mix of integers (i.e. URNs) and floats (the null is expressed as a float).

removal of on="URN",

join() defaults to using the index: we want to explicitly join in the index and by this point, can guarantee that we want to join both datasets on their respective indexes.

Checklist (add/remove as appropriate)

• Work items have been linked (use AB#)
• Your code builds clean without any errors or warnings
• You have run all unit/integration tests and they pass
• Your branch has been rebased onto main
• You have tested by running locally
• You have reviewed with UX/Design

[Faizan-ahmad00]

looks good to me.