This makes it so we don't have to try to parse the geometry JSON and
instead can just pass pre-calculated h3 values, which take fewer bytes,
everywhere.
This requires a bunch of changes since the types on these tables are
much closer to the underlying types of the Fieldseeker data we are
getting back from the API.
I now need to use proper UUID types everywhere, which means I had to
modify the bob gen config to consistently use google UUID, my UUID
library of choice.
I also had to add the organization_id to all the fieldseeker tables
since we rely on them existing for some of our compound queries.
There were some changes to the API type signatures to get things to
build. I may yet regret those.
I have to checkpoint this because I'm trying to get a very complicated
multi-layered SQL query for inserting version history into the database
and I need to improve it iteratively
I've got a new binary that I can use to directly test complex stored
procedures. This is to shorted my testing loop.
This is an intermediate step between shifting from the old fs_* prefixed
table names to an entire fieldseeker schema. At this point we have both,
and we aren't doing much with the new schema but compiling.
