… on for discrete case

…uto-handling with defaults)

… spaces

and allow coefficient to be modified, adding an informative docstring
(previous implementation was reasonable only for continuous action spaces)

Adjust parametrisation to match procedural example in atari_sac_hl

… in NPG and TRPO

 * Parameter optim must not include the actor parameters (as they are updated
   via natural gradients that are computed internally)
 * Fix incorrect optimizer instantiation in high-level API

…to running the trainer,

yet recently introduced parameter `reset_prior_to_run` of `run` suggested that it was optional.
But it was not respected, because `__iter__` would always call `reset(reset_collectors=True, reset_buffer=False)`
regardless. The parameter was removed; instead, the parameters of `run` now mirror the parameters of `reset`,
and the implicit `reset` call in `__iter__` was removed.
This aligns with upcoming changes in Tianshou v2.0.0.

responsible for creating the snapshot(s) on the original branch and then
compare with results on a modified branch.

Add writing of a log file for determinism tests.

…iner

Fix some broken tests that directly used the trainer's iterator instead of using run():
  * test/continuous/test_ppo
  * test/continuous/test_td3

…QN, QRDQN, Rainbow)

  * Require only core message equivalence (network parameter hashes)
    for the test to pass
  * Allow to ignore certain messages on a per-test level

Slightly enhanced docstrings in collector

The mechanism introduced in v1.1.0 was completely revised:
  - The `train_seed` and `test_seed` attributes were removed from `SamplingConfig`.
    Instead, the seeds are derived from the seed defined in `ExperimentConfig`.
  - Seed attributes of `EnvFactory` classes were removed.
    Instead, seeds are passed to methods of `EnvFactory`.

Control validation enabling with global flag

Don't mutate incoming dict, don't load invalid fields

Seriously improves performance of Batch constructor