Document the LDLT output representation

[sylvesterkaczmarek]

Summary

• document that the LDLT D output is an N x N diagonal matrix with zero off-diagonal elements
• define pp as an N-element sequence of row and column swaps
• explain the order required to reconstruct the permutation matrix P
• correct the public header parameter name from p to pp

Context

Issue #119 identifies two separate concerns: the square representation of D and the lack of usable documentation for the permutation vector. Changing D to a vector would alter the public API. This change preserves the current API and makes both existing output representations explicit, including the exact procedure for constructing P.

This addresses the documentation portion of #119.

Validation

• git diff --check
• compiled the scalar f32 and f64 LDLT implementations with GCC 13.3.0 and -Wall -Wextra -Werror, using minimal test-only stubs for external CMSIS Core definitions
• ran a focused f32/f64 test that reconstructed P from pp, verified P A P^T = L D L^T, and checked that every off-diagonal element of D is zero
• confirmed that the documented reconstruction order matches the existing f32 and f64 repository tests

Documentation-only change. The full repository test suite was not run.