Feature/poly eval workspace

[will-saunders-ukaea]

Bary Interpolation and polynomial basis evaluation.

Adds:

• Evaluation of Nektar++ fields using Bary interpolation and basis function evaluation (eModified_A/B)
• Projection onto Nektar++ fields using basis evaluation (eModified A/B)
• Correctness tests for basis function evaluations

Passes new unit tests and existing integration tests using (tested with intel toolchain)

[will-saunders-ukaea]

* All of the integration tests pass, except for Electrostatic2D3V.TwoStream, which mostly fails with errors like
  `The difference between c1 and 7.255197456936871 is 1.6648960104856494, which exceeds 0.73, where c1 evaluates to 5.590301446451222`
  ... although it does seem to pass intermittently with intel

I think this is caused by (not great) detection of where to fit the straight line for the energy. With the previous implementation the instability started reliably at the same place each time.

[will-saunders-ukaea]

@oparry-ukaea I think I've addressed your comments.

[oparry-ukaea]

* All of the integration tests pass, except for Electrostatic2D3V.TwoStream, which mostly fails with errors like
  `The difference between c1 and 7.255197456936871 is 1.6648960104856494, which exceeds 0.73, where c1 evaluates to 5.590301446451222`
  ... although it does seem to pass intermittently with intel

I think this is caused by (not great) detection of where to fit the straight line for the energy. With the previous implementation the instability started reliably at the same place each time.

For lack of a better idea - can we just increase the test tolerance? It would be a pain to have this keep failing

[will-saunders-ukaea]

@oparry-ukaea Can you try this commit - added some straight line detection and shifted the time interval the fit is over to slightly later.

[oparry-ukaea]

Trying it now I'm using --gtest_filter=Electrostatic2D3V.TwoStream --gtest_repeat=10 so it might take a while...!