Optimize redundant storage calls during send(_nft)_transfer

[Farhad-Shabani]

Description

While reviewing the send_transfer entrypoint, I've noticed we're making the same storage calls multiple times across different stages, which is inefficient. Specifically, this happens in the following parts of the process for the channel end:

• 1st: send_transfer_validate
• 2nd: send_packet_validate
• 3rd: send_transfer_execute
• 4th: send_packet_execute

And for the next sequence send:

• 1st: send_transfer_validate
• 2nd: send_packet_validate
• 3rd: send_transfer_execute
• 4th: send_packet_execute

Given that this is a hot path for every ICS-20 and ICS-721 packets, it's crucial to devise a more efficient implementation to optimize these redundant calls. This is the case for the send_nft_transfer() as well.