Take age into account when ranking questions

[jonhoo]

A reasonable place to start is probably https://www.evanmiller.org/ranking-news-items-with-upvotes.html

Since we currently do sorting in the client

wewerewondering/client/src/List.svelte

Lines 191 to 193 in 9c7ca7b

	qs.sort((a, b) => {
	return b.votes - a.votes;
	});

and on the server (for local operation)

wewerewondering/server/src/list.rs

Lines 69 to 77 in 9c7ca7b

	qs.sort_unstable_by_key(|qid| {
	std::cmp::Reverse(
	questions[qid]["votes"]
	.as_n()
	.expect("votes is always set")
	.parse::<usize>()
	.expect("votes are always numbers"),
	)
	});

and in DynamoDB (for prod operation)

wewerewondering/server/src/list.rs

Line 33 in 9c7ca7b

.index_name("top")

this is going to have to be repeated in a few places. DynamoDB also can't have an index on this kind of computed property, so we'll want to sort after fetching the list from DynamoDB (which will let us unify sorting between local and prod). We'll also need to expose the time a question was asked in the non-key fields in the top index that we query from so that we get it back from the query to top:

wewerewondering/infra/dynamodb.tf

Line 47 in 9c7ca7b

non_key_attributes = ["answered", "hidden"]

[jonhoo]

Why do we even sort in both places? Feels like we should be able to get rid of at least one.

Also, if the qid is a ULID, we can just grab the time out of that. Difference in minutes, rounded up, clamped to 1.. is probably fine.