Add shank thickness to the Probe model

[h-mayorquin]

Thinking about double-sided probes on probeinterface_library surfaced a gap: front and back contacts sit at identical (x, y), so the only physical separation between a front site and its back twin is the shank thickness along the plane normal, and the Probe model has no field for it. The value is published for our main manufacturers (Cambridge NeuroTech 15 microns, https://www.cambridgeneurotech.com/neural-probes; Neuropixels 24 microns, https://www.neuropixels.org/probe1-0), yet it lives only on datasheets. Even SpikeGLX's geometry metadata stores the in-plane electrode coordinates (x, z) and omits thickness (https://billkarsh.github.io/SpikeGLX/help/imroTables/).

Useful beyond double-sided probes:

• Forward-modeling of extracellular signals: the shank is an insulating body that shapes the recorded field, and ground-truth simulators model the probe body, where thickness is a direct input.
• Completeness: it is the same datasheet-sourced geometry as the contact positions and contour we already store, so the probe description stays self-contained.
• Enables reconstructing double-sided probes in three dimensions (front and back at plus or minus thickness/2 along the normal). [This is how I started thinknig about it]

Related: #14.

Altought maybe this is omited for a good reason?