fix: align ellipsis operator precedence with elixir

[rma97]

Based on #81 and fixes the failing tests.

• Treat ... as a standalone term before binary operators (like ranges).
• Allow ... to consume unary +/- so arithmetic precedence matches Elixir.
• Tests now pass for ellipsis operator precedence cases.

There were 3 failing tests:

  1) test ... operator - unary ellipsis as term before range (Spitfire.OperatorsTest)
     test/operators_test.exs:1038
     Assertion with == failed
     code:  assert spitfire_parse(code) == s2q(code)
     left:  {:error, :parse_error}
     right: {:ok, {:.., [line: 1, column: 5], [{:..., [line: 1, column: 1], []}, 1]}}
     stacktrace:
       test/operators_test.exs:1040: (test)

  2) test ... operator - unary ellipsis as term before infix operator (Spitfire.OperatorsTest)
     test/operators_test.exs:1033
     Assertion with == failed
     code:  assert spitfire_parse(code) == s2q(code)
     left:  {:error, :parse_error}
     right: {:ok, {:*, [line: 1, column: 5], [{:..., [line: 1, column: 1], []}, 1]}}
     stacktrace:
       test/operators_test.exs:1035: (test)


  3) test ... operator - unary = has higher precedence than ... (Spitfire.OperatorsTest)
     test/operators_test.exs:1018
     Assertion with == failed
     code:  assert spitfire_parse(code) == s2q(code)
     left:  {:error, :parse_error}
     right: {
              :ok,
              {:=, [line: 1, column: 5], [{:..., [line: 1, column: 1], []}, {:=, [line: 1, column: 9], [{:a, [line: 1, column: 7], nil}, {:b, [line: 1, column: 11], nil}]}]}
            }
     stacktrace:
       test/operators_test.exs:1020: (test)

[mhanberg]

Please remove the actual test.

Once we merge this, I'll rebase that PR and merge it separately.

But I can see currently that it passes so that's sweet 😎

[rma97]

Please remove the actual test.

Once we merge this, I'll rebase that PR and merge it separately.

But I can see currently that it passes so that's sweet 😎

Done