Fix the following cases:

* `build_sdist` is called with another sdist already present
  in the destination directory

* `build_wheel` is called with the destination directory not
  already created

Fix error with PEP 517 builds when wheel exists (GH #1761)

As a mitigation for #1730, this commit limits the number of workers in
the ProcessPoolExecutor to 1 (default is the number of CPUs). On PyPy,
having a higher number of available workers dramatically increases the
number of concurrent processes, leading to some resource exhaustion
issues.

This does not address the root issue, but should improve the situation
until the root issue is addressed.

`build_meta.build_wheel` assumes that the only wheel in its output
directory is the one it builds, but prior to this, it also used the
`dist/` folder as its working output directory. This commit uses a
temporary directory instead, preventing an error that was triggered when
previously-generated wheel files were still sitting in `dist/`.

Fixes GH #1671

Currently, this will fail because setuptools.build_meta.build_wheel
assumes that no wheels already exist in the `dist/` directory.

See GH #1671

This fixes GH #1682 by porting the pkg_resources requirement parsing
logic into setuptools.build_meta, so that all valid requirement
specifiers passed to setup_requires will be added to the
get_requires_for_build_* function outputs.

Fixes GH #1682

Per GH #1682, with setuptools.build_meta we are not properly handling
the situation where setup_requires is actually a newline-delimited
string rather than a list, which is supported by setup.py interface.

This adds several failing (and some passing) tests for how
setup_requires is handled by setuptools.build_meta.

Many projects can get away with an empty `setup.py` and use *only* the
declarative `setup.cfg`. With the new PEP 517 backend, we can supply a
default empty `setup.py` if one is not provided.