Clean up unused and singleton symbols in base class definitions

[rayosborn]

Closes #1659.

Implements recommendations 1 and 2 from issue #1659, which identified inconsistencies in the use of <symbols> blocks across NeXus base classes. The purpose of a symbols block is to assert that multiple fields share identically-sized dimensions; symbols that are dead (never referenced in any <dim> element) or singletons (used in exactly one field) do not serve that purpose and are removed.

Recommendation 1 — Dead symbols (zero uses in any <dim> element):

• NXsample_component: remove entire <symbols> block (n_Temp, n_eField, n_mField, n_pField, n_sField were never referenced)
• NXdetector_channel: remove dataRank and nP; strip rank="dataRank" from the two <dimensions> blocks that used it
• NXem_ebsd: remove n_op and n_z
• NXem_eds: remove n_elements

Recommendation 2 — Singleton symbols (each used in exactly one <dim>):

• NXsample: remove n_Temp, n_eField, n_mField, n_pField, n_sField; replace <dimensions> blocks with prose; n_comp (17+ uses) is retained
• NXcylindrical_geometry: remove entire <symbols> block (i, j, k); retain fixed second-dimension constraints
• NXoff_geometry: remove i, k, l; also remove the undeclared j used in winding_order; retain fixed second-dimension constraints
• NXbeam: remove c; second dimension of incident_beam_divergence left unconstrained with updated doc; nP and m (multi-use) retained
• NXelectronanalyzer: remove nfa and nsa; drop <dimensions> from fast_axes and slow_axes
• NXcrystal: remove n_comp; retain the fixed <dim index="2" value="6"/> constraint on unit_cell; i (4 uses) retained
• NXreflections: remove m; update experiments field doc
• NXapm_instrument: remove p; update signal_amplitude field doc
• NXapm_reconstruction: remove n; retain the fixed <dim index="2" value="3"/> constraint on reconstructed_positions
• NXapm_charge_state_analysis: remove n_ivec_max and n_variable; second dimension of nuclide_hash left unconstrained
• NXdisk_chopper: remove n; document the 2×n_slits length of slit_edges in prose

Recommendation 3 (moving mode-specific symbols such as nP/i/j/k/tof in NXdetector and NXbeam to the relevant application definitions) is deferred to a follow-up PR pending NIAC review.

[woutdenolf]

In addition, this PR allows reusing symbols in derived and child classes: #1650.

[woutdenolf]

The warning in nexusformat comes from the fixed rank in base classes that are not respected in the HDF5 files. Giving symbols to dimensions in base class fields does not hurt and might be used in application definitions to reference certain dimensions of fields in the base classes (used in fields defined in the application definition itself) without having to override those fields for the sole purpose of defining symbols. Of course when there are symbol collisions you will have the override those symbols in the application definition.