add torch eval_act

add package setup

add weight & bias support for logging

add git has tracker in logger

add<initialization of basic bias for basic bias cont policy>

fix<modify seed update for vecenvs && add access of attrs for vecenvs>

update some args

fix<base final activation>

add<log std clamp for policy with basic bias>

fix<squeeze action for subproc env>

fix<more ppo log & fix value loss option>

fix<more collector log>

fix<obsnormalizer snapshot>

fix<fillin snap networks for all algo>

fix<ppo in separate> && fix<collector terminate>