v3: rethink timeouts #782
Description
From comments in the read_rows PR:
nit: maybe rename to attempt_timeout?
think the params for scans should be distinct from simple unary timeouts. I think maybe having:
- operation_timeout/attempt_timeout control mutate_row, check_and_mutate_row, read_modify_write, read_row
- scan_operation_timeout/scan_attempt_timeout for read_rowS
- batch_operation_timeout/attempt_operation_timeout for mutate_rowS
Can end users disable timeouts?
Timeouts must be propagated to mutation_batcher in an intelligent way as well