Fix interpretation of 'nan(number)' by the idl_parser (#5810)

* Parser reject "nan(n)" string as it does with nan(n)

* Adjust scalar fuzzer to ignore '$schema' substrings

- Scalar fuzzer ignores '$schema' substrings at the input
- Added 'scalar_debug' target to simplify research of fuzzed cases

* Improve formatting of './tests/fuzzer/CMakeLists.txt'

tests/fuzzer/scalar_debug.cpp

@@ -0,0 +1,28 @@
+#include <iostream>
+#include "flatbuffers/util.h"
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size);
+
+int main(int argc, char *argv[]) {
+  if (argc < 2) {
+    std::cerr << "Usage: scalar_debug <path to fuzzer crash file>\n";
+    return 0;
+  }
+  std::string crash_file_name(argv[1]);
+  std::string crash_file_data;
+  auto done =
+      flatbuffers::LoadFile(crash_file_name.c_str(), true, &crash_file_data);
+  if (!done) {
+    std::cerr << "Can not load file: '" << crash_file_name << "'";
+    return -1;
+  }
+  if (crash_file_data.size() < 3) {
+    std::cerr << "Invalid file data: '" << crash_file_data << "'";
+    return -2;
+  }
+  auto rc = LLVMFuzzerTestOneInput(
+      reinterpret_cast<const uint8_t *>(crash_file_data.data()),
+      crash_file_data.size());
+  std::cout << "LLVMFuzzerTestOneInput finished with code " << rc;
+  return rc;
+}