Extract code fragment for unpacking into method in `sym_inline_unpack()`

Use `mrb_assert()` for checking `sym` in `sym_inline_unpack()`

Add `const` to `pack_table` in `src/symbol.c`

Lazy message/diff creation for assertion in `test/assert.rb`

Include the following changes too:

- Extract part of logic with block to a method
- Use `var ||= ...` instead of `var = ... unless var`
- Unify using parentheses for `t_print`
- Change methods order

Always through `assert_true` for assertion methods in `test/assert.rb`

Refine `mrbgems/mruby-bin-config/mrbgem.rake`

- Use `MRuby::Gem::Specification`
- Fix a binary name is added to `MRuby::Build#bins` multiple times
- Close file immediately (avoid using `open(...).read`)
- Simplify

Small symbols with all alphanumeric characters (<5) are packed in 32bit
symbol integer a la base64. This means those small symbols are not
listed in `Symbol.all_symbols`.

Small refactoring in `codegen.c`

NULL safed `mrb_open_core()`

Add `NoMethodError` tests to `mruby-enumerator`

* Update doc comment to clarify `Enumerator.new` without a block is
  deprecated and left only for internal use.

* Fixed some cases `mruby` raise `ArgumentError` too eagerly compared
  with `CRuby`.

* Update doc comment to clarify `Enumerator.new` without a block is
  deprecated and left only for internal use.

* Fixed some cases `mruby` raise `ArgumentError` too eagerly compared
  with `CRuby`.

Merge branch 'fix-enumerator-initialize-for-nil-or-false' of https://github.com/shuujii/mruby into shuujii-fix-enumerator-initialize-for-nil-or-false

Remove no meaning statements in `mruby-io` tests

So that `lambda{}.call(1)` raises `ArgumentError` as CRuby does.
Also, fixed junk assignment for `lambda{|;a|p a}.call{}`.

Merge branch 'rename-bin-mruby-config-to-bin-config' of https://github.com/shuujii/mruby into shuujii-rename-bin-mruby-config-to-bin-config

Integrate definition of `MRuby::Build#exefile`

In 4174e02, we removed the symbol hash table from `mrb_state` but
`find_symbol` was too slow with linear search. My performance estimation
was wrong. So we implemented a new compact hash table for symbols.