TST: test_left_eigen_vec's multi-distribution branch is never executed

Problem

Test_markovchain_stationary_distributions_KMRMarkovMatrix2.test_left_eigen_vec branches on self.n_stat_dists, but the class's only fixture is the KMR sequential-move matrix with N = 27. That chain is irreducible, so it has exactly one stationary distribution: self.stationary is always (1, 28) and n_stat_dists is always 1. The else branch at quantecon/markov/tests/test_core.py:230-233 therefore never executes.

The same applies to setup_method. stat_shape always has length 2, so the if len(stat_shape) == 1 arm at line 201 is dead too, and n_stat_dists is only ever assigned from stat_shape[0].

Nothing is broken here — the assertion that does run is the correct one for an irreducible chain. The gap is that the reducible, multiple-stationary-distribution case the else branch was written for is not covered anywhere in this class, so it is untested code masquerading as a test.

Noticed while reviewing #798, which rewrites both np.dot calls in this method to @. That PR is a faithful mechanical rewrite and the dead branch predates it, so this is a separate pre-existing gap rather than anything introduced there.

Verification

Against e7b86e5:

Suggested fix

Add a reducible fixture so the branch actually runs. A four-state chain with two absorbing states yields n_stat_dists == 2:

P = np.array([[1.0, 0.0, 0.0, 0.0],
              [0.0, 1.0, 0.0, 0.0],
              [0.3, 0.3, 0.4, 0.0],
              [0.0, 0.2, 0.0, 0.8]])

Both stationary rows satisfy v @ P == v, so the existing assertion passes unchanged once the branch is reachable — the branch is correct, it is simply never entered.

The alternative, if the multi-distribution case is judged to be covered well enough by test_markovchain_pmatrices, is to delete the else branch and the dead setup_method arm instead. Either resolution is fine; leaving an unreachable assertion in place is the outcome worth avoiding.

Acceptance criteria

• The else branch of test_left_eigen_vec is either exercised by a reducible fixture or removed
• The dead if len(stat_shape) == 1 arm in setup_method is resolved the same way
• Full suite still passes

Noticed during review of #798 (AI-assisted; claims verified against e7b86e5 on 2026-09-10).

Start in quantecon/markov/tests/test_core.py, especially setup_method and Test_markovchain_stationary_distributions_KMRMarkovMatrix2.test_left_eigen_vec. Check how the existing KMR fixture determines stat_shape and n_stat_dists, then use the proposed reducible fixture or remove the unreachable branches. Run the full test suite and confirm both acceptance criteria are resolved.