Describe the problem

In a conversation with @nvanbenschoten and @AlexTalks, we concluded that the concept of reservations could likely be removed from the lock table; instead, we could just treat the first transactional queued writer as the reservation holder (without actually naming the concept as such). Doing so would allow us to get rid of some complex state transitions that happen when trying to acquiring reservations/reservation breaking -- these involve moving requests between the queuedWriters list and the reservation, potentially back and forth. This issue serves as a placeholder to explore that idea (maybe we missed something), and if viable, to actually do so.

Additional Context

Related to Nathan's comments on #102973.

Jira issue: CRDB-27962