ptodsl: support Python and/or short-circuit in AST rewrite (#1332)

[jimmychou0]

Summary

Support Python and / or between PTODSL runtime predicates in @pto.jit / @pto.func AST-rewritten kernels (issue #1332).

Python and/or are not overloadable: evaluating a and b forces a truth check that calls __bool__ on a PTODSL runtime value during tracing and raises. This PR lowers ast.BoolOp expressions into device-side shor-circuit scf.if regions that keep Python short-circuit semantics: the RHS is only traced inside the branch that Python would actually evaluate.

Implementation

• ptodsl/_ast_rewrite.py: new _BoolOpRewriter pass that rewrites every ast.BoolOp into a right-nested lazy helper call (a and b and c -> pto._short_circuit_and(a, lambda: pto._short_circuit_and(b, lambda: c))). Covers assignments, call arguments, return, and if/while conditions.
• ptodsl/_control_flow.py: internal _short_circuit_and / _short_circuit_or helpers built on the existing pto.if_ + br.assign machinery, producing a result-bearing scf.if. The LHS is the control condition (integer operands use non-zero truthiness); the branch merge reuses the existing i1 reconciliation rules; Python bool literals materialize as i1 constants (flag and True).
• Plain Python operands (lists, tuples, strings, ...) keep native Python truthiness at trace time (e.g. TileOps templates use loops or None where loops is a Python list); floating-point control values raise a clear diagnostic.
• ptodsl/pto.py: export the two internal helpers.
• No backend changes: existing scf.if / comparisons already express the semantics.

Tests

• New ptodsl/tests/test_issue_1332_boolop_short_circuit.py covering: the issue's original and/or kernels, 3+ operand chains, RHS with division/load inside the guarded region, and/or in if conditions / native while tests / call arguments / return, static short-circuit (False and rhs, True or rhs never traced), runtime integer LHS with i1 merge, f and True bool-literal materialization, and diagnostics (float control, incompatible branch merge, ast_rewrite=False, source-less functions).
• Docs: new Short-circuit and / or section in ptodsl/docs/user_guide/05-control-flow.md with a compile-mode example, plus a summary-table row.

Validation

• Full fresh build (ptoas + PTOPythonModules) and the complete ptodsl/tests suite on dev-481211 (aarch64 Linux / LLVM 19 / Python 3.10): 42/42 pass, including test_jit_compile, test_scf_while_ast, test_issue_1126_bool_merge, all tilelib suites, and the new issue-1332 test.

Fixes #1332.