this has been bothering me for MONTHS, and it compiles faster now. also fixed some warnings

This reverts commit f116e648.

(also got rid of the map templater, which had version problems with
mustache — it's not likely anyone will need it anyways)

(meant for use of this server without exneuland / the whole map
resolution api on a dump simple workadventure backend)

turns out eitherDecodeFile' doesn't have the semantics I thought it did
(who writes functions returning either that can still fail??)

this works surprisingly well, actually

otherwise workadventure & the deployment got confused

(actually i feel like there should be an easteregg or sth for maps that
really pass everything, but i can't think of something fitting)

before this, the server would fetch updates but then not update the
corresponding reference, leading to the initial state being checked out
and linted again.

This includes the backlink to the lobby (auto-generated only for now)
and a "help!"-button for sending mails. Also general info regarding
which commit was linted / published.

i kept being confused by things happening that really shouldn't 🙈

(this also adds general cli arguments, but the main point is that
downloading monstrously large repositories is a silly idea when i'm on a
train with surprisingly bad wifi)

(though apparently there's no way to set the -LC option? would look much
nicer with that …)

(i.e. we want to still have a valid version of the map if new results
where introduced)

this removes:
 - the bbb properties
 - all explicit mentions of rc3
 - the weird script domain hacks (done via a substitution now)
 - some (few) of the weirder code choices

it also adds some more type level witchery to deal with configs, which
for some reason seems to be the hardest problem of this entire program …

also the server now does inter-assembly dependency checking!

(though this time it doesn't seem to do much, if anything)

bootstrap is horrible and should be replaced

whoops, forgot this earlier

turns out aeson really REALLY likes to keep huge scientific numbers
around, which is great if your data structures consist largely of arrays
of (small) integers!