Fix interpreter eval regressions (JPERL_EVAL_USE_INTERPRETER=1)

[fglock]

Summary

Fixes regressions in the bytecode interpreter's eval STRING functionality compared to the JVM compiler. All changes ensure the interpreter produces identical results to the JVM compiler when JPERL_EVAL_USE_INTERPRETER=1 is set.

Fixes

CompileAssignment.java

• chop/chomp used as lvalues now throw Can't modify chop in scalar assignment (matches JVM compiler)
• Global scalar assignment ( = 'x') now returns the global variable lvalue (not the rhs copy), so ($_ = 'x') =~ s/// correctly modifies $_ in-place

CompileBinaryOperator.java

• Added !~ s///r and !~ y///r error checks matching JVM EmitRegex.handleNotBindRegex
• For =~/!~ in the generic fall-through path, force SCALAR context so CALL_SUB (e.g. use constant quire => qr/.../) emits ctx=SCALAR and returns a scalar

CompileOperator.java

• Added <> diamond operator support via READLINE opcode routing to DiamondIO
• replaceRegex/matchRegex bound string (element from =~ binding) compiled in SCALAR context
• Standalone m// (no =~ binding) now matches against $_ instead of returning the regex object unused

Opcodes.java

• Added LOAD_VSTRING (337): loads v-string literals with VSTRING type, so require v5.5.640 works in eval STRING
• Renamed LIST_TO_SCALAR → LIST_TO_COUNT (count semantics, opcode 60)
• Added new LIST_TO_SCALAR (338): returns last element (correct Perl list-in-scalar semantics)

BytecodeInterpreter.java

• Wired up LOAD_VSTRING and new LIST_TO_SCALAR opcodes
• HASH_KEYS/HASH_VALUES: call .keys()/.values() on RuntimeBase instead of blindly casting to RuntimeHash, so keys undef throws the proper Perl error

OpcodeHandlerExtended.java

• executeReadline: routes non-glob scalar filehandles to DiamondIO.readline for the <> operator

InterpretedCode.java

• Disassembler updated for LIST_TO_COUNT, LIST_TO_SCALAR, and LOAD_VSTRING

Tests fixed

• op/chop.t: 137/137 ✅
• re/subst.t: 183/183 ✅ (was 182)
• re/substT.t: 183/183 ✅
• re/subst_wamp.t: 183/183 ✅
• op/tr.t: 277/277 ✅
• op/dor.t: 29/29 ✅
• op/ver.t: 54/54 ✅
• op/smartkve.t: 6/6 ✅ (was 4)
• op/index.t: 414/415 (improved from 413)