Add DDM grammar productivity checker with correctness proofs

[julesmt]

Decides whether each declared category in a dialect admits a finite parse tree. Productivity is rooted in frameworkAtomicCategories — the categories the parser constructs directly — with all other categories derived from operator chains.

• Strata/DDM/Analysis/Productivity.lean — algorithm + wrapper
• Strata/DDM/Analysis/Productivity/Spec.lean — sound, complete, and correctness proofs for the fixpoint kernel and the top-level wrapper
• StrataTest/DDM/Analysis/ProductivityTest.lean — unit tests
• Scripts/ProductivityCheck.lean — lake exe productivityCheck

One sorry remains: transitiveImports.fuel_suffices — proving the n³ fuel bound is enough for the dialect-import-graph traversal.

Issue #, if available:

Description of changes:

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[joehendrix]

Hi, this is an well written contribution — the proof work is clean and well-tested.

However, I'd recommend substantial changes for the goal of avoiding stack overflows. I may have used the wrong term when mentioning a productivity checker. The property we need is guaranteed progress — every cycle in the parser's call graph must consume input before recursing. Productivity (every category has a finite derivation) is a weaker property that doesn't prevent parser loops. For example, A → B ...; B → A ... will spin the Pratt parser without consuming input even though both categories are productive.

Checking progress requires knowing the operator syntax layout (which tokens come before recursive calls), which the OpInfo extraction discards. So this isn't something that can be adapted without a fundamental rework of the approach.

We could potentially adapt this as a DDM linter for unproductive non-terminals. They are potentially useful for dialects intended to be imported (like Init), but effectively dead code in terminal dialects.

[julesmt]

@joehendrix I took into account your 3 first suggestions. For the 4th one, It's my fault, I should have asked you just to be sure. Indeed this does not catch loops in parsing. I started working on that, not all is going to waste for the productivity checker in the sense of progress guarantee !