Extend pySpecs translator to handle all service specification constructs

[joehendrix]

Summary

Extend the Strata pySpecs Python-to-spec translator to handle the full
range of assertion patterns found across service specification files.
Before this change the translator emitted warnings on many common
patterns; after, it handles all constructs without warnings.

Builds on the robustness and API improvements from #512.

Details

• F-string messages. Assertion messages that use Python f-strings
  are now preserved semantically (distinguishing literal text from
  interpolated expressions) rather than being flattened to a single
  string.
• For-loop quantifiers. for item in kwargs["field"] loops are
  translated to forallList quantifiers. The element type is inferred
  from typing.List or typing.Sequence annotations in the
  corresponding TypedDict. for k, v in dict.items() tuple unpacking
  is handled similarly via forallDict, with key/value types inferred
  from Dict or Mapping annotations.
• Conditional implication. When assertions appear inside an if
  block (e.g. if "Items" in kwargs:), the condition becomes an
  antecedent: implies(condition, assertion). Else-branch assertions
  receive the negated condition.
• __init__ as class fields. Rather than treating __init__ as a
  regular method, the translator now extracts
  self.field = self._ClassName() assignments as typed class fields
  with an optional constant value. Unhandled __init__ patterns
  emit warnings.
• Type-checked numeric comparisons. Comparison assertions
  (>=, <=) now use type inference to choose between integer and
  floating-point operators. When a float-typed field is compared
  against an integer literal (e.g. SampleSize >= 0), the bound is
  promoted to float automatically. Negative integer bounds (>= -1)
  are also supported.
• Negative-integer DDM encoding. Fixed the negSucc round-trip in
  the DDM serialization layer so that -1 renders as "-1" rather
  than "-0".
• DDM operator precedences. Subscript, comparison, and implication
  operators have explicit precedences so the pretty-printer no longer
  emits unnecessary parentheses.
• Test coverage. Positive tests cover f-strings, for-loops, dict
  iteration, conditionals, __init__ fields, float/int comparisons,
  and negative bounds. A separate negative-test file (warnings.py)
  verifies that unsupported patterns produce the expected warnings.
  Test assertions now throw descriptive errors instead of panicking.

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

[joehendrix]

Review fixes (force-pushed)

Addressed all 13 issues from code review, rebased on latest main, and squashed into a single commit.

Correctness

• Fix missing s! string interpolation in 3 "{dialect} already loaded" error messages
• Emit warning when transCondition drops unrecognized if-conditions (previously silent)
• Fix negInt 0 DDM edge case: renamed negSuccInt to negInt, explicit 0 handling, added round-trip #guard tests

Safety

• Replace assert! with specWarning in For handler (type_comment, for/else) and class-body FunctionDef (type_comment, type_params)

Structure

• Remove dead Pred type (with stale FIXME) and Iterable type
• Extract makeComparison helper to deduplicate GtE/LtE comparison arms (~30 lines removed)
• Extract resolveBaseClasses helper to deduplicate base-class resolution in pySpecClassBody and translate

Test Coverage

• Add else-branch test exercising the not() DDM path end-to-end
• Add for_else_loop warning test using previously unused LoopRequest TypedDict
• Add negative float bound test (>= -0.5) exercising extractFloatBound with UnaryOp(USub, ConFloat)
• Warning test now verifies signatures are produced (not just warnings emitted)

Style

• Wrap 2 long lines in DDM.lean to stay under 100 characters