Eliminate double inverts in pyrtl.optimize

[gaborszita]

Closes #447

Example:

Block before optimize	Block after optimize (without my changes)	Block after optimize (with my changes)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.0%. Comparing base (5077c24) to head (4306a00).
Report is 21 commits behind head on development.

Additional details and impacted files

@@              Coverage Diff              @@
##           development    #462     +/-   ##
=============================================
+ Coverage         91.7%   92.0%   +0.3%     
=============================================
  Files               24      24             
  Lines             6403    6470     +67     
=============================================
+ Hits              5866    5946     +80     
+ Misses             537     524     -13     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[fdxmw]

Thank you for fixing this issue! A couple high-level comments before we get into coding details:

1. O(N^2) nested for loops over block.logic should generally be avoided in passes because they don't scale well as the number of nets increases - the pass will take too long to run. It seems possible to iterate over block.logic once and build a list of all the inverter chains in the block (see how other passes create a _ProducerList for example). Then I think we can iterate over that list and decide if each inverter chain can be simplified?
2. This really needs a test :) Some test cases that come to mind:
   1. Chains of 1, 2, 3, and 4 inverters (even numbers of inverters should be removed, and odd numbers of inverters should be simplified to one inverter)
   2. A chain of 2 inverters with another user of the first inverter's output (should not be simplified)

[gaborszita]

Thank you for fixing this issue! A couple high-level comments before we get into coding details:

1. O(N^2) nested for loops over block.logic should generally be avoided in passes because they don't scale well as the number of nets increases - the pass will take too long to run. It seems possible to iterate over block.logic once and build a list of all the inverter chains in the block (see how other passes create a _ProducerList for example). Then I think we can iterate over that list and decide if each inverter chain can be simplified?

2. This really needs a test :) Some test cases that come to mind:

   1. Chains of 1, 2, 3, and 4 inverters (even numbers of inverters should be removed, and odd numbers of inverters should be simplified to one inverter)
   2. A chain of 2 inverters with another user of the first inverter's output (should not be simplified)

All addressed

[fdxmw]

Please start with the comments on line 109 and 81 first. If those can be addressed, they will probably make many other comments obsolete. Thank you for improving this code!

[gaborszita]

I will fix these next week, I was busy this weekend.

[fdxmw]

I will fix these next week, I was busy this weekend.

Sounds great, no rush! Let me know if you have any questions about my comments, happy to discuss.

[fdxmw]

Overall structure of this pull request looks great, so we are on to fine details at this point. We're almost there, thank you for your patience!

[gaborszita]

Sorry for not squashing my commits, I forgot I already had commits locally when I pushed.

[fdxmw]

This should be the last round of comments, thank you for your patience!

[fdxmw]

Looks great, thank you for all your effort on this pull request!