[BUG] ActionScaling and FlattenAction treat NestedKey tuples as sequences of keys

[YeonwooSung]

Describe the bug

ActionScaling and FlattenAction wrap in_keys_inv with:

if not isinstance(in_keys_inv, (list, tuple)):
    in_keys_inv = [in_keys_inv]

A NestedKey ("agent", "action") is a tuple, so it is treated as two keys "agent" and "action".

• ActionScaling then raises ValueError (single action key per instance).
• FlattenAction silently registers "agent" and "action" as two keys.

Existing nested-key tests pass a list (in_keys_inv=[("agent", "action")]) and miss this. LastAction in the same file uses isinstance(value, (str, tuple)) and treats a NestedKey tuple as one key.

from torchrl.envs.transforms import ActionScaling, FlattenAction

ActionScaling(in_keys_inv=("agent", "action"))
# ValueError: ActionScaling only supports a single action key per instance.

t = FlattenAction(first_dim=-2, last_dim=-1, in_keys_inv=("agent", "action"))
t.in_keys_inv
# ['agent', 'action']

Expected behavior

A NestedKey tuple is one key. ActionScaling(in_keys_inv=("agent", "action")) and FlattenAction(..., in_keys_inv=("agent", "action")) should store [("agent", "action")].

Checklist

• I have checked that there is no similar issue in the repo (required)
• I have read the documentation (required)
• I have provided a minimal working example to reproduce the bug (required)

[YeonwooSung]

/assign