Move to new qadence

[inafergra]

Makes qadence-libs compatible with the new qadence, in particular changing how the old TransformedModule was dealt with. In the QNG optimizer we need to distinguish the "circuit" variational parameters (those params that affect the final quantum state of the ciruict, e.g. gate params or FM params) and "non-circuit" variational parameters (e.g. the parameters modifying the observable, which do no change the final quantum state). This is because only circuit params can be optimized with the QNG optimizer (we can only compute the QFI matrix for circuit params). Before, it was easier to distinguish both as non-circuit parameters corresponded to the TransformedModule scale and shifting, but now some new logic was needed to distinguish both.

Apart from adding this new logic, I refactored the code inside qng.py so that the different approximations have now a functional API form, more in line with torch style, and also extended the documentation a bit.

[jpmoutinho]

Thanks @inafergra. Could you bump the pyproject to 0.1.3 so we can release?