feat: Change to local Newell's normal

[jafranc]

This PR draft a proposal for switching from Newell's normal method based on global coordinates to local, i.e. from
\mathbf{n} = \sum_i P_i x P_(i+1)%n to \mathbf{n} = \sum_i (P_i - P_0) x (P_(i+1)%n - P_0).

This could avoid some precision loss when dealing with large coordinates.

Tests are also more complete.

[dkachuma]

Looks good to me and indeed should have been included in the original implementation. 2280

[castelletto1]

@jafranc Please confirm that the failing integrated tests are behaving as expected, and verify the new baselines so we can proceed with merging.

[jafranc]

@jafranc Please confirm that the failing integrated tests are behaving as expected, and verify the new baselines so we can proceed with merging.

I am looking into that

[jafranc]

@jafranc Please confirm that the failing integrated tests are behaving as expected, and verify the new baselines so we can proceed with merging.

I am looking into that

Most of the failed test are exhibiting 1e-9 diff, which I would expect but the deadOilEgg:

********************************************************************************
Error: /Problem/Solvers/compositionalMultiphaseWell/wellControls7/dCurrentBHP
	Arrays of types float64 and float64 have 4 values of which 1 fail both the relative and absolute tests.
		Max absolute difference is at index (np.int64(2),): value = -98.10000000000002, base_value = -0.0
		Max relative difference is at index (np.int64(2),): value = -98.10000000000002, base_value = -0.0
	Statistics of the q values greater than 1.0 defined by absolute tolerance: N = 1
		max = 98100000.00000003, mean = 24525000.000000007, std = 42478546.05562673
		max is at index (np.int64(2),), value = -98.10000000000002, base_value = -0.0
	Statistics of the q values greater than 1.0 defined by relative tolerance: N = 0
********************************************************************************
********************************************************************************
Error: /Problem/Solvers/compositionalMultiphaseWell/wellControls7/dCurrentTotalVolumetricRate
	Arrays of types float64 and float64 have 5 values of which 1 fail both the relative and absolute tests.
		Max absolute difference is at index (np.int64(1),): value = 1.6560123574322675e-06, base_value = -7.86199250195236e-06
		Max relative difference is at index (np.int64(1),): value = 1.6560123574322675e-06, base_value = -7.86199250195236e-06
	Statistics of the q values greater than 1.0 defined by absolute tolerance: N = 1
		max = 9.518004859384629, mean = 1.9036009718769258, std = 3.8072019437538516
		max is at index (np.int64(1),), value = 1.6560123574322675e-06, base_value = -7.86199250195236e-06
	Statistics of the q values greater than 1.0 defined by relative tolerance: N = 0
********************************************************************************
********************************************************************************

strangly enough when compared proposed and develop normals:

  normal ( 0 1 0)                                                                   | normal ( 0 1 -5.78069e-16)                                                        
 normal ( -1 0 0)                                                                  | normal ( -1 0 0)                                                                  
 normal ( 0 0 -1)                                                                  | normal ( 0 0 -1)                                                                  
 normal ( 0 1 0)                                                                   | normal ( 0 1 -1.90883e-15)                                                        
 normal ( 1 0 0)                                                                   | normal ( 1 0 -2.40862e-15)                                                        
 normal ( 0 0 -1)                                                                  | normal ( 0 0 -1)                                                                  
 normal ( 0 1 0)                                                                   | normal ( 0 1 -2.75787e-15)                                                        
 normal ( 1 0 0)                                                                   | normal ( 1 0 0)                                                                   
 normal ( 0 0 -1)                                                                  | normal ( 0 0 -1)                                                                  
 normal ( 0 1 0)                                                                   | normal ( 0 1 2.6555e-15)                                                          
 normal ( 1 0 0)                                                                   | normal ( 1 0 -4.81724e-15)                                                        
 normal ( 0 0 -1)                                                                  | normal ( 0 0 -1)                                                                  
 normal ( 0 1 0)                                                                   | normal ( 0 1 6.6237e-15)                                                          
 normal ( 1 0 0)                                                                   | normal ( 1 0 9.63448e-15)                                                         
 normal ( 0 0 -1)                                                                  | normal ( 0 0 -1)                                                                  
 normal ( 0 1 0)                                                                   | normal ( 0 1 9.38759e-15)                                                         
 normal ( 1 0 0)                                                                   | normal ( 1 0 9.63448e-15)   

so I am looking into upwinding or a normal conditions that could flip

[jafranc]

After discussion with @tjb-ltk, the discrepancy in baseline seems normally introduced by change of formula. I'll update baselines

[jafranc]

@OmarDuran @rrsettgast @jhuang2601 @bd713 can you have a look at it ?

[jhuang2601]

Suggested change

EXPECT_LT( abs( faceCenter[0]-EXPECTED_CENTER[0] ), 1e-6 );

This is the duplicate of the above line