Request to decouple PerfCounterIO's increment width from RetireWidth

Sorry for probably introducing the original hack, but I'd like to make a request to decouple the PerfCounterIO event's width from the retireWidth.

https://github.com/ucb-bar/rocket-chip/blob/master/src/main/scala/rocket/CSR.scala#L152

I would like a RocketCoreParams field that will let me choose what I want the event increment width to be.

My use-case is that a processor's issueWidth is likely to be wider than its retireWidth. Separating out different issue ports to separate events in not tenable, since software may want to save/restore counter state on a fine granularity. Even for coarse performance analysis on FPGAs, 29 hpm counters can run out fast.

Does this seem reasonable?

Start at src/main/scala/rocket/CSR.scala#L152 and trace how PerfCounterIO's event width is derived from retireWidth and how RocketCoreParams is used. Define the configurable increment-width behavior independently of retireWidth, then verify that the issue's wider-issue-width use case and counter save/restore granularity are supported.

Written by the indexing model from the issue text.