Resolve dictionary premises with bounded, context-sensitive search - #6272

Stack from ghstack (oldest at bottom):

• Keep elaboration decisions from the final name-resolution pass #6303
• Accept dictionary elaboration only after core verification #6302
• Resolve source givens from their completed binding schemes #6301
• Preserve source given scope across dependency grouping #6300
• Record completed given schemes for deferred resolution #6299
• Build complete dictionary plans independently of note order #6298
• Introduce dictionary parameters while checking qualified definitions #6297
• Order generated dictionary parameters within source scopes #6296
• Capture lexical givens by their active binder identities #6295
• Record dictionary goals at precise checker application sites #6294
• Require determined goal types before dictionary search #6293
• Insert dictionaries at identified expression sites #6292
• Identify lexical dictionary bindings without changing source diagnostics #6291
• Track expression identity separately from source location #6290
• Check implicit arrows without erasing dictionary requirements #6289
• Keep effect rows outside implicit constraint contexts #6288
• Track given membership in namespace metadata #6287
• Verify explicit dictionary bodies before accepting elaboration #6286
• Decompose qualified signatures into dictionary candidates #6285
• Handle implicit parameters in function-type utilities #6276
• Represent and store implicit function arrows #6275
• Bound the size of generated implicit goals #6274
• Prefer strictly more specific dictionary conclusions #6273
• -> Resolve dictionary premises with bounded, context-sensitive search #6272
• Match polymorphic dictionaries without solving goal variables #6271
• Select lexical dictionaries by binding identity and scope #6270
• Preserve reserved namespace names when parsing text #6269

Resolve each candidate premise after matching its conclusion to the determined goal. Require every premise variable to be fixed by that match, and freshen candidate variables before substitution. Search lexical scopes from inner to outer, prune only equal active goals, and report depth or work exhaustion as inconclusive.

Do not cache winners across different search contexts. Regression tests reject inconsistent premises and the prototype counterexample under all 720 candidate orderings, and accept the valid rigid-variable chain. Namespace aliases count as one candidate.

Validation: all 13 unit suites, 22 CLI integration tests, all 391 transcripts, round-trip and manual-rewrite checks, compiled-code and documentation checks, and formatting pass.