feat!: optimize and rework send() function

[Shvandre]

Issue

Closes #1649
Closes #1558

Checklist

• I have updated CHANGELOG.md
• I have run all the tests locally and no test failure was reported
• I have run the linter, formatter and spellchecker
• I did not do unrelated and/or undiscussed refactorings

Results (two last rows):

[anton-trunov]

Super-awesome stuff 🎸 Let's update the Jetton benchmarks

[anton-trunov]

I think all those inline comments make the code harder to read. We should

• review the functional equivalence of the old and new code
• make only high-level comments only explaining the approach, if needed

[novusnota]

I think all those inline comments make the code harder to read. We should

• review the functional equivalence of the old and new code
• make only high-level comments only explaining the approach, if needed

I've edited the context.tact part, see: 736f532. Feel free to comment on that. During this hour I'll edit the send.tact part and address possible comments for context.tact.

By the way, we also need a note on Context.readForwardFee() now working only for the workchains, not the masterchain! I'll add that in.

[novusnota]

By the way, we also need a note on Context.readForwardFee() now working only for the workchains, not the masterchain! I'll add that in.

UPD: Value of first_frac is the same for masterchain and other workchains as of now, so I was technically wrong there. For now.

---

Now that I think about it, we're (potentially) making a breaking change here compared to the non-asm version of the function: #1650 (comment)

Instead of adding the ZERO GETORIGINALFWDFEE by the end of Context.readForwardFee(), we should probably keep the fwd_fee value that it gives, and only then apply the getOriginalFwdFee() function on top.

Otherwise:

1. getOriginalFwdFee() is rendered useless, and the example given there gives higher fees than before
2. Context.readForwardFee() returns not the fwd_fee * 3 / 2 as it used to do (for whatever reason), but the (fwd_fee * 2^16) / first_frac, where first_frac is a value listed in config param 25 of the Blockchain: https://tonscan.org/config#25.

P.S.: But I may be totally off here, so I'll write some tests :)

[Shvandre]

I actually left previous behaviour

[novusnota]

I actually left previous behaviour

Yeah, and I overlooked it in the previous PR. The behavior prior to all asm changes was:

extends fun readForwardFee(self: Context): Int {
    let sc: Slice = self.raw;
    sc.loadAddress(); // Skip destination
    sc.loadCoins(); // Skip value
    sc.skipBits(1); // Skip extra currency collection
    sc.loadCoins(); // Skip ihr_fee
    return (sc.loadCoins() * 3) / 2;
}

Which, roughly speaking, was producing halved result of the current iteration of .readForwardFee(). Or at least it seems that way to me, assuming TON Docs are correct.

[anton-trunov]

@novusnota can you please suggest tests to see if your considerations are taken into account here

[novusnota]

Hmm, I did some local checks and @Shvandre is right, things are still ok. Sorry everyone, false alarm from me :):

That said, I'll add e2e tests to encode the behavior