[Feature Request] `TransformedEnv` of a `TransformedEnv` should just append the transforms, not create a new `TransformedEnv`

[vmoens]

Motivation

TransformedEnv is the TorchRL way of transforming an environment input/output.
When calling TransformedEnv(TransformedEnv(env, t1), t2) we are actually doing the same thing as TransformedEnv(env, [t1, t2]), but in a more convoluted way.
A nice feature to have would be that TransformedEnv(TransformedEnv(env, t1), t2) just appends t2 to the list of transforms (like append_transform does).