fix(perf): perf improvements for secator

[ocervell]

We want to make performance improvements for Secator: it seems we often use too much memory and our live command processing is too slow

For instance, when running a mapcidr task on a big cidr range, the task generates thousands of Ip output types and our memory usage grows enormously. On top of that, running mapcidr without secator takes a few thousands milliseconds and more than 5 seconds with secator processing.

Find performance bottlenecks (especially in our Command (runners/command.py) and Runner (runners/_base.py) classes, and propose a draft implementation to fix them.