refactor: simplify watchmap traversal

[baszalmstra]

This PR cleans up the code by adding a new type for traversing and modifying the watchmap. The propagation loop contained a number of bookkeeping locals which distract from the actual propagation algorithm. There were also methods to manipulate the watches and the watchmap in different places, they are now all consolidated in the WatchMapCursor struct.

Using the cursor is very simple: WatchMap::cursor is used to create a cursor to iterate the linked list of watches. WatchMapCursor::next moves the cursor to the next clause watching a certain literal. WatchMapCursor::update modifies the current clause/watch to watch a different literal. This method performs all the bookkeeping of the linked lists to ensure consistency.

Hopefully this makes the code easier to understand.

[baszalmstra]

Mm the benchmarks are not favorable:

main_timings.csv: 1000 records
timings.csv: 1000 records

Summary for main_timings.csv:
- Average Solve Time: 0.83 seconds (mean of all durations)
- Median Solve Time: 0.35 seconds (middle value when sorted)
- Standard Deviation: 2.95 seconds (spread of durations around the mean)
- Minimum Solve Time: 0.00 seconds (shortest solve duration)
- Maximum Solve Time: 50.00 seconds (longest solve duration, capped at 50s)
- 25th Percentile: 0.11 seconds (25% of solves were faster than this)
- 75th Percentile: 0.89 seconds (75% of solves were faster than this)
- Number of Outliers: 3 (solves capped at 50s)

Summary for timings.csv:
- Average Solve Time: 0.94 seconds (mean of all durations)
- Median Solve Time: 0.40 seconds (middle value when sorted)
- Standard Deviation: 3.06 seconds (spread of durations around the mean)
- Minimum Solve Time: 0.00 seconds (shortest solve duration)
- Maximum Solve Time: 50.00 seconds (longest solve duration, capped at 50s)
- 25th Percentile: 0.12 seconds (25% of solves were faster than this)
- 75th Percentile: 0.98 seconds (75% of solves were faster than this)
- Number of Outliers: 3 (solves capped at 50s)

Comparison between the datasets:
- Average Solve Time: 'timings.csv' was 0.88 times faster than 'main_timings.csv'
- Median Solve Time: 'timings.csv' was 0.88 times faster than 'main_timings.csv'
- 25th Percentile: 'timings.csv' was 0.93 times faster than 'main_timings.csv'
- 75th Percentile: 'timings.csv' was 0.91 times faster than 'main_timings.csv'
- Outliers: 'timings.csv' had 0 fewer solves capped at 50s

[baszalmstra]

I simplified the code even further and added a little bit of unsafe magic and the benchmarks look much better now:

main_timings.csv: 1000 records
timings.csv: 1000 records

Summary for main_timings.csv:
- Average Solve Time: 0.83 seconds (mean of all durations)
- Median Solve Time: 0.35 seconds (middle value when sorted)
- Standard Deviation: 2.95 seconds (spread of durations around the mean)
- Minimum Solve Time: 0.00 seconds (shortest solve duration)
- Maximum Solve Time: 50.00 seconds (longest solve duration, capped at 50s)
- 25th Percentile: 0.11 seconds (25% of solves were faster than this)
- 75th Percentile: 0.89 seconds (75% of solves were faster than this)
- Number of Outliers: 3 (solves capped at 50s)

Summary for timings.csv:
- Average Solve Time: 0.86 seconds (mean of all durations)
- Median Solve Time: 0.37 seconds (middle value when sorted)
- Standard Deviation: 3.00 seconds (spread of durations around the mean)
- Minimum Solve Time: 0.00 seconds (shortest solve duration)
- Maximum Solve Time: 50.00 seconds (longest solve duration, capped at 50s)
- 25th Percentile: 0.11 seconds (25% of solves were faster than this)
- 75th Percentile: 0.89 seconds (75% of solves were faster than this)
- Number of Outliers: 3 (solves capped at 50s)

Comparison between the datasets:
- Average Solve Time: 'timings.csv' was 0.96 times faster than 'main_timings.csv'
- Median Solve Time: 'timings.csv' was 0.95 times faster than 'main_timings.csv'
- 25th Percentile: 'timings.csv' was 1.01 times faster than 'main_timings.csv'
- 75th Percentile: 'timings.csv' was 0.99 times faster than 'main_timings.csv'
- Outliers: 'timings.csv' had 0 fewer solves capped at 50s

I feel this is good enough to merge.