[vouch score] refactor walk to find all simple paths

[sasuke0787]

• refactor so that we walk from target back to roots
• add improved clear test cases for branching patterns
• patch infinite loop issue not aborting on max processed
• refactor early exits for performance

[0o-de-lally]

For future reference here's the result of some benchmarking with production data:

## forward walk
#### from roots to target

| Depth |   Score   | Max Depth Reached | Accounts Processed |
|------:|----------:|------------------:|-------------------:|
|     2 |         0 |                 2 |                 28 |
|     3 |   100000  |                 3 |                 87 |
|     4 |   250000  |                 4 |                324 |
|     5 |   587500  |                 5 |               1368 |
|     6 | EXECUTION_LIMIT_REACHED |    |                    |

---

## reverse walk
#### from target to roots of trust

| Depth |   Score   | Max Depth Reached | Accounts Processed |
|------:|----------:|------------------:|-------------------:|
|     2 |         0 |                 2 |                  4 |
|     3 |   100000  |                 3 |                 12 |
|     4 |   250000  |                 4 |                 49 |
|     5 |   562500  |                 5 |                217 |
|     6 |   981250  |                 6 |               1235 |
|     7 | EXECUTION_LIMIT_REACHED |    - |                  - |

[sirouk]

LGTM 👍

[0o-de-lally]

For completness, we also tested a breadth first approach. While the numbers come out the same in the tests, with production data it's quick to exceed memory limits. This is what we would expect from a BFS vs DFS. Though it's possible it can be revisited in the future and tuned.

https://github.com/0o-de-lally/libra-framework/blob/breadth-first/framework/libra-framework/sources/ol_sources/tests/page_rank_bfs.test.move#L116