[Feature]: Full image input support

[typoverflow]

Prerequisites

• I have tried updating UtilsRL to the newest version.
• I have checked both open and closed issues but found nothing related to my request.

UtilsRL verison when proposing this request

0.3.13

What I am expecting

1. add CNN backend for UtilsRL.
2. refactor replay buffers to cope with high-dimensional visual observation.
3. (possibly) refactor replay buffers to store discrete action. This may involve a significant change to the current buffer design, for we cannot set the dtype of action to float anymore. maybe relevant to [Feature]: Choice of float number precision #2.

Possible solutions

No response

Any additional messages which might help

No response

Urgency

Urgent, will bring significant improvement and should be considered in next main version.