feat: pool.recycleWorkers() method

[AriPerkkio]

• Closes Feature: task level isolation #62

[AriPerkkio]

@sheremet-va could you test if this fixes your PR on Vitest's side?

I quickly tested this with my reproduction case from vitest-dev/vitest#3659 and it seems to work nicely. No perf issues noticable.

        // always run environments isolated between each other
        for (const env of envs) {
+         if (envs.indexOf(env) !== 0)
+           await pool.recycleWorkers()

[sheremet-va]

+         if (envs.indexOf(env) !== 0)

Shouldn't it always recycle? There is a possibility that some tests were running in multithreaded, so there are already some workers running (if you have a workspace project with isolate: false and another with isolate: true)

We also need a test for this. I think we already do, but looks like it's not correct

[AriPerkkio]

Shouldn't it always recycle?

My reproduction case was simple enough that it did not create conflicts there. But sure, it's OK to set pool to recycle when ever needed. It's not that slow to create new set of workers.

[sheremet-va]

@sheremet-va could you test if this fixes your PR on Vitest's side?

Can confirm that tests in vitest-dev/vitest#3491 are passing

[AriPerkkio]

We also need a test for this. I think we already do, but looks like it's not correct

vitest-dev/vitest@7d8413a