Eliminate the separate lexer module and its dynamic token allocation by moving token types, lexer state, and parsing functions directly into compiler.c. The compiler now receives a raw source string and length, lexes tokens on the fly into a stack-allocated Token struct, and discards the Buffer abstraction, token linked list, and associated memory management routines.
Replace the recursive-descent parser that built an AST (Node* tree) with a single-pass compiler that emits bytecode directly into a Block structure. Remove the parser.c/parser.h files and their AST node types (NodeLiteral, NodeCall, NodeBinaryOp, etc.), add new compiler.c/compiler.h with a Compiler struct and Pratt-style precedence parsing that outputs CODE_CONSTANT and CODE_END instructions, and introduce vm.c/vm.h containing a Fiber-based interpreter with a stack, interpret() loop, and printValue() for OBJ_INT. Update main.c to call compile() then interpret() instead of parse(), remove TOKEN_INDENT/TOKEN_OUTDENT from token.h and dumpTokens(), and adjust the Xcode project to include the new source files while excluding the old parser.
