chore: new tket-qsim unitary simulator for unit testing

[ferbetanzo]

Closes #1648. Includes the following:

• New tket-qsim crate that allows to simulate (purely unitary) hugrs. It defines the required <tket|helios|sol>_ops unitary matrices but compares them against roqoqo's gate definitions.
• New semantics preserving tests for cross-compilation.

[codecov]

Codecov Report

❌ Patch coverage is 85.86207% with 123 lines in your changes missing coverage. Please review.
✅ Project coverage is 86.24%. Comparing base (7e114dd) to head (5bde2bf).

Files with missing lines	Patch %	Lines
tket-qsim/src/simulate.rs	76.19%	54 Missing and 16 partials ⚠️
tket-qsim/src/matrix.rs	86.06%	28 Missing ⚠️
tket-qsim/src/tket_ops.rs	96.20%	9 Missing ⚠️
tket-qsim/src/helios_ops.rs	87.09%	8 Missing ⚠️
tket-qsim/src/sol_ops.rs	89.47%	8 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1886      +/-   ##
==========================================
- Coverage   86.25%   86.24%   -0.02%     
==========================================
  Files         202      207       +5     
  Lines       31992    32862     +870     
  Branches    30500    31370     +870     
==========================================
+ Hits        27594    28341     +747     
- Misses       3165     3272     +107     
- Partials     1233     1249      +16     

Flag	Coverage Δ
python	92.18% <ø> (ø)
qis-compiler	74.17% <ø> (ø)
rust	85.95% <85.86%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 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.

[doug-q]

Thanks for this. Apologies but I do not think we should proceed with this approach. Happy to discuss.

I think that we can't use selene because of dependency loops, is that right?

Problems with this approach

• It relys on the hugrs under test to get optimised into single-dfg functions
• lowerings that contain control flow cannot be handled
• lowerings that include ancillas or mid circuit measurements cannot be handled
• The hugrs that are tested are fixed. This approach is not going to catch corner cases (e.g. bugs where special angles cause big problems)

I don't think this is the right solution to verifying our transformations.

I suggest

• Each op should be verified independently
• The expected unitary should be specified local to the test
• I don't think we should build an interpreter here. (I would be ok on depending on a separate, complete(ish) interpreter)
• A better approach, IMO, would be to
  • depend on selene in such a way as it doesn't create circular dependencies. This is surely possible because we don't need to run any tket transformations in selene (we can do it beforehand)
  • For parameterised gates use property testing to explore the parameter space and catch corner cases

[doug-q]

I would prefer to use the https://docs.rs/approx/latest/approx/ crate for this

[doug-q]

I would prefer to use the https://www.nalgebra.rs/docs/ crate for this