Add parallel compilation of programs sent by the client

[Tsygankov-Slava]

Current behavior:

If first client sends code A, which runs ~5000ms, and the second client sends code B, which runs ~100ms, then a queue is formed and the second client will receive it's output only after executing code A, that is after ~5100ms.

Expected behavior:

If first client sends code A, which runs ~5000ms, and the second client sends code B, which runs ~100ms, then the requests should be parallelized and the first client should receive a response in ~5000ms, and the second in ~100ms.