ENH: Add state_values to DiscreteDP

[HG-Cheng]

This PR partially addresses #248 by adding state_values support to
DiscreteDP.

Changes:

• Add an optional state_values argument to DiscreteDP.
• Validate state_values consistently with MarkovChain.
• Preserve state_values when converting between product and
  state-action-pair formulations.
• Pass state_values to the MarkovChain returned by controlled_mc.
• Add tests for storage, validation, propagation, and representation
  conversion.

action_values and num_actions are intentionally left out of this PR,
following the review discussion.

[coveralls]

coverage: 90.57% (+0.02%) from 90.546% — HG-Cheng:feature-ddp-values into QuantEcon:main

[oyamad]

@HG-Cheng Thank you for the contribution!

Main comments

1. I propose that we only add state_values and remove action_values (and num_actions) in this PR.

   Please see Adding State-Action Pair Support QuantEcon.jl#94 (comment):
   We are implicit which of the two approaches in that comment we are following (although I believe approach 1 is the legitimate approach). We need a discussion to decide on which approach to pick (or even whether to pick one explicitly). I propose not to discuss it here, focusing only on adding state_values in this PR.

2. Please pass state_values=state_values to MarkovChain in controlled_mc: https://github.com/HG-Cheng/QuantEcon.py/blob/fff5405ae7d4f66df60bb2ae39a943504be19c59/quantecon/markov/ddp.py#L1004
   This is where adding state_values is most useful.

Other comments

Please do not include changes in the files .vscode/settings.json and _matrix_eqn.py in this PR.

[HG-Cheng]

"Hi @oyamad, thanks for the detailed review and guidance!

I completely understand. I will remove action_values and num_actions to keep this PR focused on state_values only, and we can leave the state-action pair discussion for the future.

I will pass state_values=state_values to MarkovChain in controlled_mc as suggested.

Apologies for the accidental inclusion of .vscode/settings.json and _matrix_eqn.py! They leaked in from my local workspace and another task. I will remove them from this branch right away.

I'll push the updated commits shortly!"

[HG-Cheng]

Thanks again for the review. I've updated the PR to keep it focused on
state_values only, removed the unrelated changes, and propagated
state_values to the controlled MarkovChain.

I also synced the branch with the latest main. The full local test suite passes locally (605 passed).

[oyamad]

Thanks @HG-Cheng! One main request: please follow what is done in MarkovChain:

• Implement state_values as a @property with a validating setter (storing to self._state_values), and have __init__ delegate to it by self.state_values = state_values, as in MarkovChain.__init__.
• You might also use MarkovChain's error messages and setter docstring verbatim (e.g., 'state_values must be an array_like of length n').

[HG-Cheng]

Thanks for the review, @oyamad!

I've updated DiscreteDP.state_values to follow the existing MarkovChain implementation:

• added a state_values property with a validating setter;
• stored the value in self._state_values;
• changed __init__ to delegate through self.state_values = state_values;
• aligned the validation messages and setter docstring with MarkovChain;
• added tests for reassignment, unsetting, and invalid assignment after construction.

The focused tests and the full local test suite pass (605 passed).

Could you please take another look when convenient? Thanks!

[oyamad]

@HG-Cheng Thanks!

@mmcky This is ready to merge.