Add transform dispatcher to programs

[rmoyard]

Description of the Change:

It makes very easy for users to create a custom TranformProgram.

program = TransformProgram()
program.add_tranform(transform)

This will be used in the device to create the pre processing program.

[github-actions]

Hello. You may have forgotten to update the changelog!
Please edit doc/releases/changelog-dev.md with:

• A one-to-two sentence description of the change. You may include a small working example for new features.
• A link back to this PR.
• Your name (or GitHub username) in the contributors section.

[mudit2812]

Looks great! Thanks Romain

[codecov]

Codecov Report

Patch coverage: 100.00% and no project coverage change.

Comparison is base (975d3f4) 99.66% compared to head (232ea70) 99.66%.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #4559   +/-   ##
=======================================
  Coverage   99.66%   99.66%           
=======================================
  Files         376      376           
  Lines       33047    33059   +12     
=======================================
+ Hits        32937    32949   +12     
  Misses        110      110           

Files Changed	Coverage Δ
pennylane/transforms/core/transform_program.py	98.95% <100.00%> (+0.14%)	⬆️

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

[timmysilv]

the order still seems strange to me, but it's clear enough. Curious about in-place changes as mentioned in comments.

I'm also wondering - you chose dispatcher.add_to_program(program) instead of program.add_dispatcher(dispatcher). The latter seems more intuitive (kinda like list.append(item) instead of item.add_to_list(list)), could you explain why? I almost feel like we can update push_back to just check if it's passed a dispatcher or a container, then do the right thing depending on that. Or just put it into a new function like push_transform or push_dispatcher (the former will make more sense to end-users)

[timmysilv]

this looks really great! to be clear, expand_transforms don't all need to be at the front of a program, just before their associated (regular) transform, right?

[rmoyard]

@timmysilv the only constrain for expand transforms is that they need to be just before applying the transform itself.

[timmysilv]

🎉