Add the tent map's four generating partitions as one family

[Autoplectic]

Summary

At the Misiurewicz point the tent map's kneading partition (cut at the critical point c = 1/2) can be refined by either or both of c's order-1 preimages, L = 1/(2a) and R = 1 - 1/(2a). That gives four generating partitions of the same dynamics. This PR defines all four together rather than treating the fully-refined one as a special case.

• One table of the four machines with uniform accessors keyed by the cuts they make — tent_map_misiurewicz_partition_{cuts,forward,symbol_matrices,information_expected} plus TENT_MAP_MISIUREWICZ_PARTITIONS. States are named by decreasing stationary weight in every partition so they are comparable across the family; for "c" that relabels the published figure, and tent_map_misiurewicz_fig7_symbol_matrices still carries the figure's own labels.
• Every machine was re-derived exactly in the number field Q(a) with a**3 = 2a + 2, so each transition probability and each ephemeral rate is a quadratic in a with rational coefficients — no a-dependent denominators.

All four are generating, hence share h_mu = log2(a), and all four are unifilar and strictly sofic with infinite Markov and cryptic order. Only the anatomy split moves:

Partition	Cuts	States	Alphabet	r_mu	r_mu ≈
"c"	c	4	2	(59 + 7a - 11a**2)/57	0.6483
"Lc"	L, c	5	3	(56 + 25a - 23a**2)/57	0.4953
"cR"	c, R	5	3	(1 - 6a + 4a**2)/19	0.1529
"LcR"	L, c, R	5	4	0	0.0000

Two exact identities came out of the derivation and are now tested:

• The rate published by James et al. (2013), (1/4)(3 - 2/(a+1) - 4/(a+2) + 9/(2a+3)), reduces modulo the minimal polynomial to (59 + 7a - 11a**2)/57 — an independent check of the whole derivation against the paper.
• The ephemeral rate is exactly modular over the two cuts: r_mu("c") - r_mu("Lc") - r_mu("cR") + r_mu("LcR") = 0 identically in a. Each cut is worth a fixed number of bits whether or not the other has been made, and the L cut's share is exactly the invariant measure of the two interval-Markov cells it separates. Excess entropy is modular too; statistical complexity is not.

Test plan

• pytest — 927 passed
• ruff check . and ruff format --check .
• ty check — no new diagnostics
• sphinx-build -W --keep-going -b html docs — no content warnings
• All four machines cross-checked against an independent chaos_anatomy float pipeline (edges, causal-state weights, h_mu/b_mu/r_mu, excess entropy) to ~1e-12
• Symbolic tests assert row sums normalize identically in a free a, and that the modularity identity and the reduction of the published formula hold symbolically