This fixes a bug that was causing cell data to not get written since
November 9th, 2023. When batch inserting cells while processing plugins,
I allowed overwriting cells that had `is_base_game = true`. Since I
always set `is_base_game = false` for cell upserts from plugins, this
was causing the base game cells to revert to `is_base_game = false`. All
it took was one mod to bundle `Skyrim.esm` for this to happen.
This broke writing cell data since `get_cell_data` depends on the
`is_base_game` value to find edits to the Skyrim base game cells.
To prevent this in the future, batch inserts when processing plugins is
no longer allowed to update cells which have `is_base_game = true`. The
only time we allow upserting these rows is when running the
`is_base_game` backfill which initially seeds the database with the base
game cells.
I had a bad unique index on the `cells` table. This commit does not fix
it but has two backfill scripts that will allow me to fixup the data
before creating the new unique index.
`is_base_game` identifies cells that are from the `Skyrim.esm` shipped
with the game (and not from some mod that happens to also have a
`Skyrim.esm` with a 0,0 cell with a different form id, see:
https://github.com/thallada/modmapper-web/issues/8).
`dump_cell_data` now only dumps cells from `is_base_game` cells.
`deduplicate_interior_cells` removes duplicate cell and plugin_cell rows
caused by the buggy index.
I will need to upgrade the database and create a new index in a later
commit.
Not adding it to the update.sh script quite yet since I need to run the very long process of downloading every classic skyrim mod that currently exists on the nexus first.
I was accidentally omitting plugins from the plugins data list when the `updated_after` parameter was used. Splitting it up into two queries makes it much simpler.
