Important adjustments to the batch function

[PaulRBerg]

Context

I've recently tweeted about our batch function, and it turned out to be a productive exercise, as we have received a lot of great feedback in the replies.

Changes

I've parsed through all replies, and I would like us to make the following changes to batch:

• Bubble up the revert (for better specificity when debugging errors, and not needing a Tenderly account to debug batch-related reverts in Sablier)
• Explain in the NatSpec comments that it is NOT safe to use msg.value; quote this article
• Return results in an array

Rationale

Why return the results?

1. They provide critical information for onchain integrators. Imagine needing to know the stream IDs that have just been created. It's impossible to do that now when batch is used.
2. The additional gas cost isn't much.

I've made a gas comparison:

• Without returned results: 257,276
• With returned results: 261,698

So it's only 4422 extra gas for creating two streams, i.e., extra ~2.2k gas per created streams. Small price to pay.

Resources

Check out PRBProxy for inspiration on how to bubble up the revert, as well as this StackExchange Q&A.

Note: unlike in PRBProxy, let's NOT revert with a custom error when the revert response is empty. Let's just bubble it up empty, as is.