i#6971: Use instr count instead of wallclock for simulated time

[derekbruening]

When using the drmemtrace scheduler in an analyzer or other tool that does not track simulated time with the default QUANTUM_INSTRUCTIONS, the scheduler used to use wall-clock time to measure blocking-input and idle time. Here we change that to use the instruction count plus the idle count via a new idle counter.

The time_units_per_us and sched_time_units_per_us defaults are set to 1000, reflecting a 2gHz machine with IPC=0.5.
The old time_units_per_us=100 for wall clock was too low; to match it with counts, we need a low sched_time_units_per_us: 500 is better than 1000, but that seems unrealisitc. Instead we can get the results we want
from our large traces by exiting earlier, since most of the unwanted idle is still in seemingly unrepresentative regions at the end. We raise exit_if_fraction_left from 0.05 to 0.1 here.

Using counters provides a more reproducible result across different runs and machines.

Wall-clock time is still used to measure idle time on replay. Switching to the idle count added here is left as separate work under #7023. (Replay also uses wall-clock time to coordinate concurrent outputs beyond shared input constraints; that will likely always remain.)

The new default values of the options were tested on larger traces and found to produce a representative level of idle time.

This change means that the clock going backward problem (#6966) is no longer seen in default runs. The analyzer still supports wall-clock with the -sched_time option so a check to avoid underflow is added.

Fixes #6971
Fixes #6966

[derekbruening]

I abandoned the idle weighting as it did not seem to have a good
justification based on evaluating idle vs instructions in schedule_stats.
I needed it because the old time_units_per_us=100 for wall clock
was too low; to match it with counts, either the time_units_per_idle
scaling is needed or dropping time_units_per_us from 1000 to 500
but that seems unrealistic. Instead we can get the results we want
from our large traces by exiting earlier, since most of the unwanted
idle is still in seemingly unrepresentative regions at the end.
The plan is to raise exit_if_fraction_left from PR #7018 to 0.25 by default.
I didn't want to delete this branch and remake it on top of that one so
will make that change after PR #7018 goes in.

[brettcoon]

As I understand it, this change only affects the QUANTUM_INSTRUCTION scheduler mode, replacing wall-clock time with instructions + idles, (except for schedule replay that still uses wall clock). Can you add something like that (or a corrected version), to the check-in comment? It wasn't initially clear to me from the existing text exactly what modes were affected.

[derekbruening]

As I understand it, this change only affects the QUANTUM_INSTRUCTION scheduler mode, replacing wall-clock time with instructions + idles, (except for schedule replay that still uses wall clock). Can you add something like that (or a corrected version), to the check-in comment? It wasn't initially clear to me from the existing text exactly what modes were affected.

Updated.