Randomgen

This release syncs most upstream changes in `Generator`. It also fixes two edge case bugs that affect `ThreeFry`, `Philox` and `AESCounter`.

This is a minor release with two issues:

* Future proof against `setuptools` changes and warnings
* Improve documentation in `RDRAND`

This is a bug-fix release with one fixed issue:

* Corrected `RDRAND` to retry on failures with pause between retries. Add a parameter ``retry`` which allows the number of retries to be set. It defaults to the Intel recommended value of 10. Also sets an exception when the number of retries has been exhausted (very unlikely). See the `RDRAND` docstring with unique considerations when using `RDRAND` that do not occur with deterministic PRNGs.

This is a minor release with two new features:

* Added `Romu` which is among the fastest available bit generators.
* Added `SFC64.weyl_increments` to simplify generating increments for use in parallel applications of `SFC64`.
* Completed QA of all bit generators to at least 4TB.

This is a feature and bug-fix release. The key new features are:

- **Deprecated** `Generator` and `RandomState` since these are both available in NumPy.
- Added `ExtendedGenerator` which contains features not in `numpy.random.Generator`.
- Added `UserBitGenerator` which allows bit generators to be written in Python or numba.
- New bit generators:
- Added `PCG64DXSM` which is is the 2.0 version of PCG64 and will likely become the default bit generator in NumPy in the near future.
- Added `EFIIX64` (a.k.a. `efiix64x48`) which is both fast and high-quality.
- Added `SFC64` which supports generating streams using distinct Weyl constants.
- Added `LCG128Mix` which supports setting the LCG multiplier, changing the output function (including support for user-defined output functions) and pre- or post-state update generation.
- Added `LXM` which generates variates using a mix of two simple, but flawed generators: a Xorshift and a 64-bit LCG. This has been proposed for including in Java.
- Tested all bit generators using at least 1TB in PractRand in different configurations. This is WIP and all generators will be tested to at least 4TB.
- Additional features and fixes:
- Added support for the `dxsm` and `dxsm-128` variants of `PCG64`. The `dxsm` variant is the official PCG 2.0 generator.
- Added support for broadcasting inputs in `ExtendedGenerator.multivariate_normal`.
- Added support for the `++` variant of `Xoroshiro128`.
- Fixed a bug the produced incorrect results in `randomgen.mt19937.MT19937.jumped`.
- Fixed multiple bugs in `Generator` that were fixed in `numpy.random.Generator`.

* Fix a build bug affecting Windows wheel generation (