refactor libvje (remove per-frame thread churn)

[game-stop]

The parallel region should be moved to vje_fx_apply()

-> keeps thread pool alive for the whole FX chain

#pragma omp parallel num_threads(global_threads)
{
        for (each effect in chain) {
            effect_apply(...); 
        }
}

serial sections must be guarded with #pragma omp single

#pragma omp single
{
    serial_setup();
}

#pragma omp for simd schedule(static)
for (...) { ... }

#pragma omp single
{
    serial_finish();
}

1. remove "parallel" from all effect-level pragmas (362 hits)
2. use #pragma omp for / #pragma omp for simd for pixel loops.
3. guard all serial code with #pragma omp single.

result:

• faster processing of FX chain

While touching the .C file:

1. review description
2. review parameter hints
3. review beat parameter hints