* Gracious quit of fast_parallel_environment

Previously, when there are a lot of parallel environments, ctrl-C often cannot terminate
the training successfully. Although "ctrl-|" can be used to terminate, it leaves the
system is a bad state which subsequent training often fails.

This PR make the ProcessEnvironment detect the KeyboardInterrupt and inform the main
process to quit by throwing an exception. This exception will be handled in python
code and thus calling FastParallelEnvironment.close() to graciouly terminate all
the processes.

* Showing progress for closing all processes

* Short message on in-algorithm checkpoint loaded

* use common.info

* implement bipartite matching loss

* fix _penv location and upgrade torch version in cicd

* fix checkpoint test case after upgrading pytorch

* faster matching loss

* remove forward()

Added the missing Boost after PR fast parallel environment

* FastParallelEnvironment

* Address comments

Also fixed a bug of not copying step and reset result.

* Address comments

* Fix unittests

* Improve setup.py

pybind11 is not always available. Add it as a prerequisite for setup in pyproject.toml

(cherry picked from commit 11318a98b839b37575052b2db233615b87cc91db)

Co-authored-by: Le Horizon <le.zhao+git@horizon.ai>

* correctly check if a thread env should be created or not

* fix format

Previously, there is the warning even if there is no unused parameter, which can be confusing.
Change to only show the warning if there are actually unused parameters.

Previously, confirm_checkpoint_upon_crash is not set False for some test (e.g ac_cart_pole),
which caused failed test waiting indefinitely.

* fix alf snapshot in grid search dirs

* fix format

* Fix save optimizer checkpoint

* Use pretty format on the structure file

* update cicd docker and alf gpu docker usage

* update alf software dependencies for cuda 11.4

* fix cicd docker file dependencies

* upgrade absl-py to be compatible with py38

* numpy version fixed to 1.23

* np.asscalar is deprecated in new numpy

* simple_value no longer works after upgrading

* skip test_mbrl_latent_pendulum

* upgrade tensorboard to 2.6.0

* upgrade cnest to 1.1.1 (fixed segfault)

* revert test case comments

* Save checkpoint structure

* Save checkpoint structure to json

* Address comments

1.1.0 uses a new packaging method, which should solve the issue of not finding .so file.

* In algorithm checkpoint pre-loading

* Explicitly call _post_init() since decorator conflicts with alf.config

* init_sub_class

* Enable auto post_init call; change to prefix@path

* Address comments

* Address more comments

* normalized action alf wrapper; transform_nest supports numeric fields

* address comments

* code style fix

* Update contributing.rst about git pull

For new version of git,  the default `git pull` will rebase, which is what we want to avoid. So we need to update the instruction.

* Update contributing.rst

* running spare env code for slow env resets, still need to figure out why reset takes 2+ seconds.

* fix pack_sequence_as(ts_spec) with ts_spec_with_env_info

* fix env_id in returned timesteps, with test

* Allow early reset even when num_spare_envs = 0, with test case.

* alf snapshot only copies alf module

* fix examples dir

* get_mode for mixture distribution

* Add comments

* Address comments

* add RandomCrop gym wrapper

* address comments

* Unify ParallelNormalProjectionNetwork and NormalProjectionNetwork

* Add make_parallel() back for normal projection network

* Add parallel FC version of make_parallel to BetaProjectionNetwork

* Implement the MixtureProjectionNetwork

* Address comments

* Unify ParallelNormalProjectionNetwork and NormalProjectionNetwork

* Add make_parallel() back for normal projection network

* Add parallel FC version of make_parallel to BetaProjectionNetwork

* VQ VAE Algorithm

* Address comments

* Fix embedding summary

* Support replace() for TensorSpec

* Add type hints for replace()