Confidence intervals from linear regression

[adrienbanse]

Following #95, for each value that is a result of a linear regression, we explicitely compute uncertainty and provide 95% confidence intervals with y = f(x) +- 1.96 * std_dev.

In the context of the computed impacts, the standard deviations for both the energy and the latency have been computed in https://github.com/genai-impact/methodology/blob/main/lab/lin_reg.ipynb. The impacts that depend on these values are therefore ValueRange, and not float or int such as before.

This PR also adds somes tests for ValueRange, as __mul__ and __truediv__ operations were implented.

Remains to

• Update the doc