We already have counters for first_open and last_close, but these additional counters will allow us to have a better understanding of how long the first open and last close operations really take.

This enhancement would also allow us to better reason about cases where some processes are still writing a shared file while others are starting to close it: without knowing when the first close operation started, we can't easily tell at which points in the application processes began to move between I/O phases.