fix: block mload merging when src and dst overlap

[charles-cooper]

What I did

fix an optimization bug introduced in #3483 when the target architecture is cancun. prior to this PR, certain sequences of mload/mstore could incorrectly be merged into mcopy when there is overlap between the source and destination buffers.

How I did it

How to verify it

Commit message

this commit fixes an optimization bug when the target architecture has
the `mcopy` instruction (i.e. `cancun` or later). the bug was introduced
in 5dc3ac7. specifically, the `merge_mload` step can incorrectly merge
`mload`/`mstore` sequences (into `mcopy`) when the source and
destination buffers overlap, and the destination buffer is "ahead of"
(i.e. greater than) the source buffer. this commit fixes the issue by
blocking the optimization in these cases, and adds unit and functional
tests demonstrating the correct behavior.

Description for the changelog

Cute Animal Picture

[chen-robert]

nit: think generally params should be "safe-by-default", and the "safer" option would be to have no overlap.

[charles-cooper]

hmm i see, my thinking was that this only affects where the source and destination address spaces are the same, and there is only one case for this, mload / mstore. but i am open to changing the default.

[chen-robert]

do you have to do do a + 32 somewhere here?

[charles-cooper]

i think the <= handles that, see the output of this "partial" overlap:

@external
def partial() -> uint256:
    a: uint256 = 1337
    b: uint256 = 0
    c: uint256 = 0
    d: uint256 = 0
    e: uint256 = 0

    c = a
    d = b
    e = c

    return e

[codecov-commenter]

Codecov Report

Merging #3635 (d996a1e) into master (c913b2d) will increase coverage by 0.09%.
Report is 1 commits behind head on master.
The diff coverage is 100.00%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

@@            Coverage Diff             @@
##           master    #3635      +/-   ##
==========================================
+ Coverage   89.01%   89.10%   +0.09%     
==========================================
  Files          86       86              
  Lines       11460    11463       +3     
  Branches     2606     2606              
==========================================
+ Hits        10201    10214      +13     
+ Misses        836      830       -6     
+ Partials      423      419       -4     

Files	Coverage Δ
vyper/ir/optimizer.py	98.31% <100.00%> (+<0.01%)	⬆️

... and 4 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[chen-robert]

lgtm