mstatus should be cached on reads.

[en-sc]

There is an issue: mstatus shares fields with other registers. See #1217.
Therefore simply caching mstatus on reads is not possible.
However, caching mstatus on reads allows to greatly reduce the number of register accesses in a typical Debug session.
The task is:

1. Measure the impact of caching mstatus on reads -- maybe it's actually not that crucial?
2. Come up with an approach that allows to cache mstatus (e.g. invalidate mstatus when accessing all the registers that mirror an mstatus field).

[JanMatCodasip]

Perhaps some form of "impact list" could be implemented for the cached registers: Invalidate the cached register value if any register on its "impact list" is written to.