Add high-level Pythonic API layer

[jamestjsp]

Summary

Wrap the 587 low-level SLICOT routines with a high-level Pythonic API module (ctrlsys/control.py) so users don't need to know SLICOT routine names or parameter conventions.

Before:

from ctrlsys import sb02od
rcond, x, alfar, alfai, beta, s, t, u, iwarn, info = sb02od(
    'D', 'G', 'N', 'U', 'Z', 'S', n, m, a, b, q, r, l
)

After:

import ctrlsys
K, X, ev = ctrlsys.lqr(A, B, Q, R)

Why

• Multiplies audience 100x — control engineers know lqr(), not sb02od
• python-control killer — same SLICOT routines MATLAB uses, but with a modern Python API
• Pure Python, zero risk — thin wrappers over existing C bindings, no build changes
• AI-agent friendly — agents call ctrlsys.lqr() without memorizing 587 routine signatures
• guide.rst already has the mapping — implementation is straightforward

Proposed functions (priority order)

Function	Description	Underlying routine(s)
lqr(A, B, Q, R)	Continuous LQR	SB02OD
dlqr(A, B, Q, R)	Discrete LQR	SB02OD (DICO='D')
lqe(A, C, Q, R)	Kalman filter gain	SB02OD
solve_care(A, B, Q, R)	Continuous algebraic Riccati	SB02MD/SB02OD
solve_dare(A, B, Q, R)	Discrete algebraic Riccati	SB02MD/SB02OD
hinf_norm(A, B, C, D)	H-infinity norm	AB13DD
h2_norm(A, B, C, D)	H2 norm	AB13BD
balred(A, B, C, nr)	Balanced truncation	AB09AD
hinf_syn(A, B, C, D)	H-infinity synthesis	SB10AD/SB10DD
pole_place(A, B, poles)	Pole placement	SB01BD
ctrb(A, B)	Controllability matrix/test	AB01MD
obsv(A, C)	Observability matrix/test	AB01MD (dual)
gram(A, B, C, type)	Gramians	SB03MD
lyap(A, Q)	Lyapunov solver	SB03MD
dlyap(A, Q)	Discrete Lyapunov	SB03MD (DICO='D')

Implementation notes

• Pure Python module, ~20-40 lines per function
• Input validation + sensible defaults for SLICOT parameters
• Return named tuples or simple tuples matching MATLAB conventions
• NumPy arrays in, NumPy arrays out (F-order handled internally)
• Errors raised as Python exceptions, not info codes