improve CI time: parallelize general state test

[jangko]

Experiment using evmstate tool to execute general state test in multiple process yield promising result.
The time needed to execute the test formerly 17m+, but after parallelized, it down to 3m+.

Using nim std lib osproc startProcess combine with chronos async works well on both linux and windows.
but the same code produce flaky result on macos. if we can isolate macos problem and fix it, it will reduce CI time significantly

[jangko]

The slowness in Windows CI have been identified and fixed fixed. No longer need for paralelization.

• Switch C compiler from GCC to llvm-mingw.
• Use libbacktrace instead of baked in nim stack strace.
• Use explicit 127.0.01 instead of 'localhost' for all tests involving network services.