Support for grazing-incidence exit angles and polar angle

[EdgarGF93]

The units called previously exit angles should be "scattering_angles" in the classic way, because they don't depend on incident and tilt angles. Rotations are applied to q units.
Now, the unit called "exit angle" is used in the grazing-incidence community for the scattering angle with the origin in the horizon of the thin film, now dependant on the incident and tilt angles
The "exit_angles" support is especially useful for SAXS experiments

[EdgarGF93]

In the first image, while incident_angle=0.0, the exit angle is 0.0 in q=0, which is also the horizon
In the second image, at iangle=0.2 rad, the exit angle is 0.0 in the horizon, the direct beam is at -iangle (-0.2 rad) and the reflected beam (at Bragg condition) is at 0.2 rad

This example is just for visualize the difference, if the images are integrated using the angle units, there is no missing wedge

[EdgarGF93]

In a real example, we would integrate using the angle units:

%matplotlib widget
incident_angle = 0.0
tilt_angle = 0.0
res2d_0 = fi.integrate2d_grazing_incidence(data=arr, 
                                           unit_ip="scattering_angle_horz", unit_oop="exit_angle", 
                                           incident_angle=incident_angle, tilt_angle=tilt_angle,
                                          )

incident_angle = 0.2
tilt_angle = 0.0
res2d_1 = fi.integrate2d_grazing_incidence(data=arr, 
                                           unit_ip="scattering_angle_horz", unit_oop="exit_angle", 
                                           incident_angle=incident_angle, tilt_angle=tilt_angle,
                                          )

fig, axes = subplots(ncols=2)
plot2d(res2d_0, ax=axes[0])
plot2d(res2d_1, ax=axes[1])

[EdgarGF93]

All these images will be generated in the tutorial notebook:

If we dont specify the units (so it will use qip_nm-1 and qoop_nm-1):

for so in range(1,9):
    ax_ = ax.ravel()[so-1]
    res = fi.integrate2d_grazing_incidence(data=data, sample_orientation=so,
                                       # unit_ip="qtot_nm^-1",
                                       # unit_oop="chigi_deg",
                                      )

We see that our "correct" sample orientation is 6 (or 7, it only switches the x sign)

for so in range(1,9):
    ax_ = ax.ravel()[so-1]
    res = fi.integrate2d_grazing_incidence(data=data, sample_orientation=so,
                                       unit_ip="qtot_nm^-1",
                                       unit_oop="chigi_deg",
                                      )

So in this case, if we choose the correct sample orientation, the polar angle 0 will be associated always to the vertical axis of the sample (fiber axis or thin film normal vector)

[EdgarGF93]

needs to fix #2389

[EdgarGF93]

Needs to add tests for the different units with scales horizontal and vertical (non-regression)