Add more strict Cell benchmarks

[elisachisari]

Currently Cell benchmarks are compared to the expected uncertainty in the measurement (10%). However, we are seeing some differences between Cells between different calls to CCL which could be affecting Fisher forecasts. For this reason, we propose having a more strict benckmarking of the Cells.

This might imply revising the decision tree for integration methods.