This enhances self-containment.

- Changed the `mrb_callinfo::pc` field to point to itself.
  Previously it indicated the return destination of the previous call level.
  `mrb_callinfo::pc` will now hold the address to its own `proc->body.irep->iseq`.
- Removed `mrb_callinfo::err` field.
  This is because `mrb_callinfo::pc - 1` is semantically the same as the previous `err`.
- The `pc0` and `pc_save` variables in `mrb_vm_exec()` are no longer needed and have been deleted.
- It removes the argument because `cipush()` doesn't need to save the previous `pc`.

This enhances self-containment.

Previously `mrb_context::stack` had the current call level stack, but now it owns it.
The `mrb_context::stack` field, which is no longer needed, will be removed.

If there is `env`, `env->c` means `target_class`.

This is because it is enough to express the range up to (-1..255) or (-3..255).

The first byte of UTF-8 character should not be `80..c1`.

Remove unnecessary `#include` in generated files

Reapply 116e128b because it is back at 456878ba

Delay test code build until `rake test`

`mrb_fiber_resume()` can be called from C; ref #3056

With this change, the test code will not be built unless `rake test` is
run, so there will be almost no side effects even if `enable_test` is
always set (but, gems specified by `add_test_dependency` are included
in `libmruby.a`).

Also added are `test: build` task, which only builds the test code
(including the main code), and `test: run` task, which only runs tests
independent of build. Therefore, the idiom for building in parallel and
not running tests in parallel is `rake -m test:build && rake test:run`.

Merge pull request #5266 from shuujii/stop-immediate-flush-to-stdout-stderr-when-running-in-parallel-on-CI

Stop immediate flush to stdout/stderr when running in parallel on CI

During parallel execution (build), it is sometimes flushed before line
breaks, and lines are concatenated.

Use namespaces according to Rake conventions

Merge pull request #5264 from shuujii/immediately-flush-output-to-stdout-stderr-so-that-logs-are-not-mixed-on-CI

Immediately flush output to stdout/stderr so that logs are not mixed on CI

Set in build configuration to enable on all CI platforms.

Change the Rake task to one that uses namespaces as follows (previous task
names can also be used for compatibility).

| Previous Task  |    New Task    |
|----------------|----------------|
| api_doc        | doc:api        |
| capi_doc       | doc:capi       |
| clean_doc      | doc:clean      |
| clean_api_doc  | doc:clean:api  |
| clean_capi_doc | doc:clean:capi |
| view_api       | doc:view:api   |
| view_capi      | doc:view:capi  |
| gitlab_config  | gitlab:config  |
| gitlab_dockers | gitlab:dockers |

Add scan target functions for presym

Add the following functions:

- mrb_intern_cstr
- mrb_define_singleton_method
- mrb_define_class_under
- mrb_define_module_under

Fix line continuations with mixed indentation in squiggly heredocs

Merge pull request #5259 from shuujii/consider-the-case-a-local-variable-name-does-not-become-a-named-presym

Consider the case a local variable name does not become a named presym

Non-ASCII characters are allowed to local variable names, so they are not
always named presym.

Fix build with `enable_cxx_exception`

Take advantage of gembox

Merge branch 'improve-source-scanning-for-presym' of https://github.com/shuujii/mruby into shuujii-improve-source-scanning-for-presym

But you still cannot cross C function boundary.