(1) Using incorrect daylight savings offset for date should warn
	Like using CDT when we are not in dalyight savings. CST would be
	correct, but maybe it was on purpose? We should not hard-error,
	just print somewhere so we see it Someday.

(2) Configurable Logging
	Learn how to use tracing better so we can mark CSS and Images
	with their file type in the traces and filter them out if we
	want to.

(3) Cleaner templates in writings
	i have apparently never tried to use a bempline thing in the
	actual content of a thing. I have learned  that this does not
	work! I had to bodge a system where I tell the formatter it's
	a template and then I throw it through a different template
	function where not all of the variables are available?
	I /really/ need to clean this up

(6) Draw lower-on-average graph line last
	Drawing the lower graph line last, for networking i am thinking
	about specifically, will allow it to be drawn over the larger
	one so it will let us see most of both fo them. This is working
	now with TX/RX because we do more TX than RX, but that could
	change and I would rather have it switch with smartness.

(7) Stick 404's in the database
	Mostly because I think it would be funny to see all the weird
	endpoints the crawlers hit, and a bit to know what links I've
	killed over the course of time.

[next todo number: 8]

# DONE

(5) Only grab a data sample every minute
	related to (4). currently we grab a datapoint on boot always,
	but that will lead to a large number of minutes having more
	than one sample, which will make the graphs end short of we
	implement (4) how it's described. we could grab more than the
	required points, but that is wasteful.
	on boot, before the first point, check when the last sample
	was taken and wait long enough to be a minute apart if needed.

	done :)
	we grab the last meminfo sample and wait until a minute after
	that one to start data collection again.

(4) Make the statistics graphs have a concept of time
	they do currently not a concept of time. all they do is pull
	the latest 256 points and graph them regardless of whether or
	not the full minute has elapsed or if more than a minute
	elapsed. this is not very good.
	get the last 256 points and correctly form them into a graph,
	not taking more than one point per minute and skipping gaps
	in data.

	done :)
	we now go through the samples and place them correctly in
	time-slots. this both removes too-many-per-minute as well
	as lets us display gaps in time.
	The type /did/ have to change from [usize; 256] to an
	[Option<usize>; 256], though. but i'd rather represent
	gaps as nothing instead of as 0, which is for-sure something.