Throw an error in writer if property isn't defined

[amc-corey-cox]

This adds an error for Koza if the writer is given properties to write that aren't defined. Previously, we silently ignored those columns and just didn't write them.

[amc-corey-cox]

I have implemented the main check. We should consider if we want to allow this feature to be turned off. I also need to write tests.

[amc-corey-cox]

I took a step back and decided to abstract the initialization and writer functionality back up into the KozaWriter parent class. I think this works pretty well. Please take a look and see what you think!

[amc-corey-cox]

The key files to look at are writer.py tsv_writer.py and jsonl_writer.py. Everything else is just to fix tests.

[glass-ships]

looks like mostly formatting changes? (eesh i never set up ruff or pre-commit on this repo did i)

[glass-ships]

i think i really like this approach! splitting write_node and write_edge out into their own functions is a good call too.

(i know i can't officially approve any PRs but i'm bored and wanted to share my thoughts :) )

[amc-corey-cox]

It just occurred to me that this has the potential to break some of our ingests. Now that we are requiring all columns sent to writer to be written, if an ingest relies on the previous behavior of ignoring extra columns it will break. I have not implemented making the error optional so if we need that I can; it would certainly be easy to do.

[glass-ships]

my initial thought was, "well it's probably worth fixing those ingests if they do have that dependency", as it seems a bit weird to want to write columns that don't exist in the data.
but an ignore flag is probably harmless enough?

[amc-corey-cox]

I just pushed another commit that better splits up and abstracts the init. I also noticed some unnecessary code I had forgotten to delete. I think this improves the previous refactor a fair amount making things even simpler and better partitioned.

I have put in hooks for ignoring the parameter checks for testing purposes but I haven't implemented them at a config level. I think we should just fix any ingests that don't have things listed properly. We can finish the implementation of ignoring the checks if we find a good use-case. Otherwise, I think it is bad practice to pass data to writer that you don't want to write and so it should remain an error.