The default implementation of `mrb_immediate_p` uses `mrb_type`. However,
in `MRB_WORD_BOXING`, `mrb_type` has many branches and is inefficient, so
provide an implementation that does not use `mrb_type`.

Merge pull request #4677 from shuujii/array-permutation-combination-without-block-should-return-self

`Array#(permutation|combination)` without block should return `self`

Remove unneeded `/src/y.tab.c` in `.gitignore`

Currently, `y.tab.c` is generated to under the `build/` directory.

Fix build of `mruby-random` on 32-bit mode

Fix `Array#sample` with `MRB_INT32`

Array index became potentially negative because `uint32_t` is cast to
`mrb_int`.

Fix wrong argument for `fprintf` in `mruby-bin-mrbc`

Remove unused `random.h`

Fixes for Doxygen docs

Defer several build libraries loading until needed

Fix `Range#max` test (`TypeError` is raised) on 32-bit mode

Suppress warning by gcc with `-Wmaybe-uninitialized`

Create a symbolic link in the temporary directory; fix #4642

Remove unused constant in `Rakefile`; ref e312842a

Because the linker gives a warning on FreeBSD 12.0.

```
warning: mktemp() possibly used unsafely; consider using mkstemp()
```

Merge pull request #4662 from shuujii/fix-RBreak-exceeding-6-words-on-32-bit-mode-w-o-boxing-and-MRB_USE_FLOAT

Fix `RBreak` exceeding 6 words on 32-bit mode w/o boxing and `MRB_USE_FLOAT`

ref: https://github.com/mruby/mruby/pull/4483#issuecomment-498001736

In this configuration, `tt` of `RBreak::val` is set into `RBreak::flags`.

Refine processing for gem lock file

- Defer YAML library and lock file loading until needed.
- Don't write empty parts into lock file.
- Extract code to read/write lock file to `MRuby::Lockfile`.
- `MRuby::Lockfile.disable` disables the use of lock file.

make clean error workaround

Take commit hash of mruby too

Separate repos directory and build directory

closes #4652

Support lock file for git.

Prioritize embedded string over nofree (or normal) string

Prioritize embedded string in the following functions:

- `str_new_static`
- `str_new`
- `mrb_str_new_capa`
- `mrb_str_pool`

The reasons are as follows:

- Consistency with `mrb_str_byte_subseq` and `str_replace`.
- Memory locality increases and may be slightly faster.
- No conversion cost to embedded string when modifying the string.

Remove deprecated iij stuffs