bootstrap: Improve CPU Utilization in SNS-Worker

[goshawk-3]

Description

The simplified scheduling approach in the current implementation means that not all CPU-cores will be fully utilized in every situations.

E.g on a machine with 32 cores, if config.batch_limit = 32, it might happen that 32 (single-block) conversions are computed sequentially where each conversion utilizes only one core.

A better approach would be to maximize core utilization. In this case, the sns-worker should convert 32 blocks in parallel thus processing 32 different (single-block) ciphertexts simultaneously.

[immortal-tofu]

We can spend time on this issue, but keep in mind that mid-term we'll run sns on GPU.