Specifically for endgames you pretty much only need to know two patterns that are rather algorithmic:
How to close a game where you’re up a rook or a queen (without stalemating).
How to close a game where you’re up one pawn, if it’s not a forced stalemate.
Two bishops is maybe something you want to learn. Theoretically knight + bishop as well. I play a lot of bullet chess and pretty much never see these though.
The rest you can really only learn with practice and pattern recognition, like how to move your pawns and how to use a minor piece to secure a win. Well, you can try to learn it algorithmically but I wouldn’t recommend it. After learning those I would move onto tactics, because those are generally how you gain a material or positional advantage that doesn’t rely on a very trivial mistake like leaving a piece uncovered.
I had to play B + N over the board early in my career, so it's definitely YMMV. It takes very little study: For B + N knowing that you mate in the corner that matches the color of the bishop, and that the knight needs to 'lead' in the effort to corral the king is all you really need to know. A half day of practice would cement it, I suspect.
I think bishop+knight endgame could be much more common if players actually tried to get them when defending after exchanges of queens and rooks. So, exchange as many pawns as possible and maybe sacrifice a piece for a couple of last pawns and here your opponent goes.
Also, if memory servers me I've seen two grandmaster games only during this week, where this endgame could happen, but the players didn't want their opponents to show the technique. One of these games was yesterday:
https://www.chessbomb.com/arena/2020-russian-championship-su...
But indeed, rook endgames are way much more important from a practical point of view.
