Landings on a large canvas no longer hold up the server

When many cards land one after another on a canvas with hundreds of archived cards, the server rebuilt every card after each single write and went through the whole archive to do so, which held it up for a large part of a second. Now a write to one card rebuilds only that card, and the database finds the cards it needs through indexes. A run of thirty landings costs about a third of the CPU it did before.