Redesign some driver commons

[alecandido]

The common Instrument and Controller interfaces are pretty restricted, but there are at least a couple of elements that should be redesigned, namely

• bounds
• sampling_rate

There may be others (e.g. giving a clear role to setup()), but at least these should be rethought, especially taking into account #918.

The main motivation is that these attributes were designed assuming a monolithic controller, which is supposed to have a single set of memories, and a single acquisition/synthesis rate.
However, even the clustered devices that we have are modular, and they may require a different processing, broke down per-module.

Actually, #918 was thought for an even more general setup, where we could not even rely on a single driver for the synchronization (thus requiring a separate trigger handling).
But this is already extremely close to the modular setups we have, that separating the instructions upload and triggering may already be beneficial for the present drivers.

[alecandido]

@stavros11 any opinion about?