Add warning for poor activities convergence

[povik]

Please consider having a warning like this.

I've found on some designs in OpenROAD-flow-scripts the propagated activities don't converge mostly due to bad features in the input netlist. Having a warning like this would help uncover these cases.

I've picked an arbitrary value of 0.20 for the max change threshold, and as for the mesage number I picked one above the existing Power.cc message. Not sure if messages.txt is maintained by some automated tooling but I tried inserting the message manually.

[akashlevy]

@povik (For future reference, etc/FindMessages.tcl regenerates doc/messages.txt)

[parallaxsw]

This warns for nearly every power regression I have even though the activities indeed do converge. Consider a circuit with no feedback. When the activities propagate on the first pass they change by whatever the input activity is (with may be > .2). Since there is no feedback they are not re-queued and regs is empty. The activities are converged but it still warns about non-convergence. I think the true criteria for non-convergence is "pass == max_activity_passes_". All of the opensta regressions I have converge, so maybe the place to start is to make a test case and submit it as an issue.