[Bug #20453] segfault in Regexp timeout

https://bugs.ruby-lang.org/issues/20228 started freeing `stk_base` to
avoid a memory leak. But `stk_base` is sometimes stack allocated (using
`xalloca`), so the free only works if the regex stack has grown enough
to hit `stack_double` (which uses `xmalloc` and `xrealloc`).

To reproduce the problem on master and 3.3.1:

```ruby
Regexp.timeout = 0.001
/^(a*)x$/ =~ "a" * 1000000 + "x"'
```

Some details about this potential fix:

`stk_base == stk_alloc` on
[init](dde99215f2/regexec.c (L1153)),
so if `stk_base != stk_alloc` we can be sure we called
[`stack_double`](dde99215f2/regexec.c (L1210))
and it's safe to free. It's also safe to free if we've
[saved](dde99215f2/regexec.c (L1187-L1189))
the stack to `msa->stack_p`, since we do the `stk_base != stk_alloc`
check before saving.

This matches the check we do inside
[`stack_double`](dde99215f2/regexec.c (L1221))

regexec.c

@@ -4218,6 +4218,7 @@ match_at(regex_t* reg, const UChar* str, const UChar* end,
 
  timeout:
   xfree(xmalloc_base);
+  if (stk_base != stk_alloc || IS_NOT_NULL(msa->stack_p))
       xfree(stk_base);
   HANDLE_REG_TIMEOUT_IN_MATCH_AT;
 }

test/ruby/test_regexp.rb

@@ -1827,6 +1827,17 @@ class TestRegexp < Test::Unit::TestCase
     end;
   end
 
+  def test_bug_20453
+    assert_separately([], "#{<<-"begin;"}\n#{<<-'end;'}")
+    begin;
+      Regexp.timeout = 0.001
+
+      assert_raise(Regexp::TimeoutError) do
+        /^(a*)x$/ =~ "a" * 1000000 + "x"
+      end
+    end;
+  end
+
   def per_instance_redos_test(global_timeout, per_instance_timeout, expected_timeout)
     assert_separately([], "#{<<-"begin;"}\n#{<<-'end;'}")
       global_timeout = #{ EnvUtil.apply_timeout_scale(global_timeout).inspect }