feat: Add Goblin Ultra Circuit builder

[ledwards2225]

Description

Initial conception of the Ultra Goblin Circuit Builder.

The Goblin builder extends the existing Ultra builder. For now I've implemented the new methods directly in the existing UltraCircuitBuilder. So essentially the UltraCircuitBuilder now is the Goblin Ultra circuit builder.

This work essentially adds three components on top of the Ultra builder: 1) An op_queue that stores the ecc operations in a raw form, e.g. {op, point, scalar}, and also computes the corresponding operations natively behind the scenes, 2) op_wires which are similar to the existing wires object but are used to store the values in the "ecc op gates", and 3) methods like queue_ecc_mul_accum() and queue_ecc_add_accum() which update both the op_queue and the op_wires accordingly.

I've included some tests that check the underlying native ECC operations in the op queue, the consistency of the values written to the op_wires by the builder, and some checks of higher level functionality like batch_mul().

There are lot's of small TODOs to be addressed in the coming weeks.

Checklist:

• I have reviewed my diff in github, line by line.
• Every change is related to the PR description.
• The branch has been merged with/rebased against the head of its merge target.
• There are no unexpected formatting changes, superfluous debug logs, or commented-out code.
• There are no circuit changes, OR a cryptographer has been assigned for review.
• New functions, classes, etc. have been documented according to the doxygen comment format. Classes and structs must have @brief describing the intended functionality.
• If existing code has been modified, such documentation has been added or updated.
• No superfluous include directives have been added.
• I have linked to any issue(s) it resolves.
• I'm happy for the PR to be merged at the reviewer's next convenience.

[Rumata888]

It's not clear what the test is supposed to test from the name

[ledwards2225]

Changed to InternalAccumulatorCorrectness

[Rumata888]

This interface forces us to always redo this computation. First you need to compute the value externally, then the same value is computed in OpQueue. What do you think of making the return value the point and having no argument to the function? Then we can check the value correctness in tests

[ledwards2225]

Changed this so that instead of calling get_accumulator() explicitly, the equality op implicitly uses used the internal accumulator and that value is returned when calling queue_ecc_eq() for use elsewhere in the circuit if necessary.

[Rumata888]

LGTM