Draft: Add momentum- and position-based pseudorapidity calculations to CompositeParticle

The CompositeParticle eta calculation uses the secondary vertex and PV positions to calculate pseudorapidity. This choice was not clear to users and doesn't give sensible results for non-displaced vertices. It could also provide worse resolution than a momentum based calculation for fully reconstructed particles.

This MR:

• Removes the CompositeParticle::eta
• Adds CompositeParticle::eta_position and CompositeParticle::eta_momentum
• Replaces eta with eta_position and eta_momentum in all selections. Unless the line authors' intentions were clear, I used eta_position for lines selecting partially reconstructed decays and eta_momentum for lines selecting fully reconstructed decays.

FYI @gligorov

device/combiners/src/FilterSvs.cu

@@ -88,8 +88,8 @@ __global__ void FilterSvs::filter_svs(
     dec = vertex.vertex().chi2() > 0 && vertex.vertex().chi2() < maxVertexChi2;
     // Kinematic cuts.
     dec &= vertex.minp() > minTrackP;
-    dec &= vertex.eta() > minChildEta;
-    dec &= vertex.eta() < maxChildEta;
+    dec &= vertex.eta_position() > minChildEta;
+    dec &= vertex.eta_position() < maxChildEta;
     dec &= vertex.dira() > minCosDira;
     // Kinematic cuts - IPchi2(dec2), IPlow(dec3) and IPhigh(dec4)
     bool dec2 = true;