feat: Implement persistent RPN state for all RPN-related input

- Modified ParseAndEvaluate() to not reset stack at beginning, only initialize if nil
- Changed runRPN() to use getRPNState() for persistent state across calls
- Fixed Stack type definition ordering in variables.go to follow Go best practices
- Updated tests to create fresh RPN instances where independence is needed

This enables REPL-style incremental RPN calculations like '3 4 +' followed by '5 +'.

1 files changed, 2 insertions, 3 deletions

diff --git a/internal/repl/repl.go b/internal/repl/repl.go
index f79f92f..3e2d4c9 100644
--- a/internal/repl/repl.go
+++ b/internal/repl/repl.go
@@ -109,9 +109,8 @@ func executor(input string) {
 
 // runRPN parses and evaluates an RPN expression
 func runRPN(input string) (string, error) {
-	vars := rpn.NewVariables()
-	rpnCalc := rpn.NewRPN(vars)
-	return rpnCalc.ParseAndEvaluate(input)
+	state := getRPNState()
+	return state.rpnCalc.ParseAndEvaluate(input)
 }
 
 // isBuiltinCommand checks if input starts with a built-in command