Hybrid histogram: bin each molecule to a node in its own compartment

[jcschaff]

Fixes #25.

The bug. VCellSmoldynOutput::computeHistogram bins each molecule to its nearest grid node, whatever compartment that node is in. Next to a curved membrane, a molecule inside the cell can be nearest to an exterior node. The PDE does not solve the coupled species there, so that molecule's contribution is lost.

The fix. The nearest same-compartment re-binning written for this in 2011 was commented out for performance (742e6d7), before the spatial hybrid existed. Since then, posincompart has gained the highResVolumeSamples map, which removes most of that cost. This PR:

• re-enables the re-binning;
• restores isInSameCompartment from 1edb089;
• fixes the node centre it compares against, which left out the mesh origin and used dy for z.

The issue has the full history.

Measured. Native runs through libvcell → pyvcell-fvsolver: ball R = 4 µm on a 37³ mesh of [0,9]³, 8 seeds.

	stock 0.10.7	#32 + this PR
A binned to nodes outside the cell (conversion)	1.23%	0.003%
conversion A_p → B: A + B mass balance at 2 s	−0.57%	+0.19%
exchange A_p ⇌ B: A + B mass balance at 5 s	−6.8%	−2.1%

• Attribution. This PR alone takes the exchange balance from −6.8% to −5.2% (measurement in Hybrid histogram bins particles to nodes outside their compartment (same-compartment correction is commented out) #25). The rest of the improvement is the creation fix in Hybrid 0th-order creation: clipped boundary cells, and cells that straddle a membrane #32.
• What remains. Most of the remaining −2.1% fits the 1.8% volume gap between Smoldyn's smooth membrane (261.8 µm³) and the voxel PDE domain (266.7 µm³), which Hybrid histogram bins particles to nodes outside their compartment (same-compartment correction is commented out) #25 discusses.

Not covered. I haven't run the C++ test suite locally; CI covers it. The cases come from viva-pde-particle, Study B2b sphere-cosim-vs-native.

🤖 Generated with Claude Code