[FEAT] Allow match_weights_chart with model json only

[RossKen]

Is your proposal related to a problem?

Currently to generate a match_weights_chart a linker (with data) is required. If I am trying to visualise a trained model which has been saved as a json, this feels unnecessary.

My current workaround is to create an empty dataframe with the correct column nomes, which isn't ideal.

Describe the solution you'd like

Alternative to linker.visualisations.match_weights_chart that is not part of the linker that allows you to pass a trained model. For example:

from splink.visualisations import match_weights_chart

chart = match_weights_chart("model.json")

Similar to the clustering methods that live outside of the linker

Describe alternatives you've considered

Additional context