Stabilize RHF energy increases and add anthracene sample

[mveril]

The anthracene/cc-pVDZ calculation repeatedly jumped by about 180 hartree while the SCF residual stayed near 40. RHF now rejects DIIS proposals that increase the energy beyond roundoff, clears the history, and minimizes the exact quadratic RHF energy along a fresh physical density step. Mixed intermediate densities cannot satisfy the convergence check; final orbitals still pass canonicalization before MP2.

DIIS also retries shorter histories when the solve is singular, nonfinite, or requires excessively large cancelling weights. The PR includes the anthracene geometry and its cc-pVDZ calculation.toml with 300 iterations, a 1e-8 threshold, six DIIS vectors, and MP2. The temporary scf-check.toml has been removed locally and is not included.

Validation:

• cargo test -p rustiq-core --no-default-features: 160 unit tests, 18 integration tests, and 4 documentation tests passed.
• cargo clippy -p rustiq-core --all-targets --no-default-features -- -D warnings
• cargo fmt --all -- --check
• Regression tests cover a deliberately uphill DIIS proposal and subsequent recovery to the stored PySCF water reference, energy minimization, electron count, Fock consistency, and unstable DIIS histories. Existing reference energies and tolerances are unchanged.

The full anthracene calculation was not rerun at the user's request. Its convergence with this correction remains unverified.