Midplane and internal-inductance entries made consistent with the DD

The load_conventions accessor was run before and after the change; both quotes are recorded on disk under the run directory.

Midplane entries

Before. Both entries fixed the midplane as the point on the separatrix contour where the tangent is vertical:

outboard_midplane equation: \left.\frac{dR}{dZ}\right|_{\mathrm{sep}} = 0,\quad R > R_{\mathrm{axis}}
inboard_midplane  equation: \left.\frac{dR}{dZ}\right|_{\mathrm{sep}} = 0,\quad R < R_{\mathrm{axis}}
symbols:  R(Z)|_{\mathrm{sep}} = the major radius of the separatrix contour as a function of height
          R_{\mathrm{axis}} = major radius of the magnetic axis

Reason. The point is undefined for a limited plasma, which has no separatrix, and it disagrees with poloidal_angle, whose zero lies on the outboard side of the horizontal plane through the magnetic axis.

After. The accessor returns:

outboard_midplane equation: Z = Z_{\mathrm{mid}},\quad R > R_{\mathrm{axis}}
inboard_midplane  equation: Z = Z_{\mathrm{mid}},\quad R < R_{\mathrm{axis}}
symbols:  Z_{\mathrm{mid}} = the midplane height the data's midplane identifier selects,
                             the magnetic-axis height when none is given
          R_{\mathrm{axis}} = major radius of the magnetic axis

The DD makes the midplane a per-dataset choice: equilibrium/time_slice/profiles_1d/midplane_identifier takes 1 magnetic_axis, 2 dr_dz_zero_sep, 3 z_zero or 4 ggd_subset. The general height Z_{\mathrm{mid}} replaces the hard-coded dr_dz_zero_sep condition.

Internal inductance

Before. The definition named no form, while the DD carries several (l_i(1), l_i(2), l_i(3), differing in the denominator radius and the leading factor).

After. The accessor returns the l_i(3) form: "The l_i(3) form of the plasma's internal inductance: the volume integral of the squared poloidal magnetic field over the plasma, normalized by the plasma current and the major radius of the plasma's geometric axis." The equation and symbols are unchanged.

Measure

Gate tests/grammar/test_conventions.py and tests/test_vocabulary_consistency_regressions.py: 20 passed, EXIT=0 at base d0b7eafb, and 20 passed, EXIT=0 at head; zero added failures.