lib/gis: solve write out conflicts of G_percent/G_progress when multithreading

[cyliang368]

This PR should solve #5776.

If we want to solve writing conflicts to the same place (stderr), it must be serial.

[petrasovaa]

Thanks! There is probably no better solution, but I think this still needs some benchmark whether this has any impact on the speed.

[marisn]

This should not be merged, as:

• it has a serious speed penalty
• it does not solve [Bug] G_percent is not safe to be called from parallel code #5776

[cyliang368]

Thanks! There is probably no better solution, but I think this still needs some benchmark whether this has any impact on the speed.

This should not be merged, as:

• it has a serious speed penalty

It is a very small part, so I believe its impact could be negligible.

I simply ran v.surf.rst and r.univar a few times manually, and I could not tell whether the difference was made by this change or some uncertainty from OS.

I would like to get some benchmarks on r.mapcalc after #5742 is merged.

---

• it does not solve [Bug] G_percent is not safe to be called from parallel code #5776

@marisn Can you elaborate on why you think it does not solve #5776?