Merge pull request #4379 from shuujii/extract-similar-codes-to-macros-for-math-opcode-in-mrb_vm_exec

Extract similar codes to macros for math opcode in `mrb_vm_exec()`

Fix memory leak for hash table index if occur out of memory

Fix wrong size of instance variable if occur out of memory

Fix leaked function symbols

Include `RFiber` and `RIstruct` as a part of `RVALUE`

Fix hexdigits convertion

- `free_heap()` in src/gc.c
- `symhash()` in src/symbol.c
- `no_optimize()` in mrbgems/mruby-compiler/core/codegen.c

`sizeof(string-literal)` is included `'\0'` character

Fix broken NaN with `MRB_NAN_BOXING`

Example:

  $ bin/mruby -e '(Float::INFINITY - Float::INFINITY).nan?'
  zsh: segmentation fault

Cause:

  `SET_FLOAT_VALUE` is not used. It is needed for normalizing NaN.

Treatment:

  In my environment, this issue could be reproduced only when
  `infinity - infinity`, however `SET_FLOAT_VALUE` should be used in all
  arithmetic operations (regardless of boxing setting), I think.

  So I fixed all similar codes by extracting to macro.

Refine `assert_float`

The `env` stores stack length in a 10 bit field.
See `MRB_ENV_STACK_LEN()` macro.

Avoid arithmetic operations when `exp` and/or `act` are infinity or NaN.

Remove incorrect flags updating in `mrb_regexp_p()`

Fix buffer overflows in parser.

Use `mrb_immediate_p()` in `mrb_obj_freeze()` and `mrb_obj_frozen()`

Extract frozen checking to function

Remove unneeded function prototypes