Any feedback/objections on introducing a new "Signatures" sysvar which would allow for a transaction's signature to be accessible inside an instruction? For reference, I created a SIMD proposal here.

This would allow for an instructions to use signatures in a few different ways:

• Store inside an account's data for a program to use later on
• Use signature in some hashing mechanism which would be use by an offline application
• Create a chain of events which use "previous" signature as a way to fetch previous entries without needing getSignaturesForAddress, etc.
• Fair instruction execution is ensured by distributing transactions into shards based on the signature bytes, using the slot modulus to determine the shard assignment
• etc..

Aside from specific ways listed here on how an instruction would utilize a signature, we should empower developers to use their imagination and build novel protocols. I don't see any downside, be it performance (99% of Tx will only have a single 64-byte signature) or technical implementation (relatively easy to implement - happy to do so), but can see many upsides listed above.

Thoughts?