[PExplicit] Improve state caching, add non-chronological backtracking

[aman-goel]

Updates include:

[State caching]

• Adds storing hash code and string representation during PValue initialization for performance
• Adds new state caching modes using hash functions from Google's guava library
• New state caching modes include: hashcode (using hashCode() from PValue and Java built-ins, 32-bit hashes), siphash24 (64-bit), murmur3_128 (128-bit, default), sha256 (256-bit), or exact (using exact string representation).

[Non-chronological backtracking]

• Adds non-chronological search that implements search beyond a fixed order (e.g., beyond depth-first or breadth-first).
• Breaks down exploration into SearchTask, where each task explores a slice of the execution tree.
• Each search task tracks a set of contiguous unexplored choices it is assigned to explore. It either explores all of them completely, or creates new children tasks for remaining unexplored sub-choices.
• Exploration within a SearchTask is chronological, exploration across SearchTasks is non-chronological and can jump at arbitrary places in the execution tree.
• New CLI option --strategy to set the search strategy: dfs, random (default), astar
• New hidden CLI option --schedules-per-task <uint> to set the number of schedules explored in each SearchTask in random/astar strategy. Default is 100. For dfs, there is no limit.
• New hidden CLI option --children-per-task <uint> (default 2). This relates with how packaging of SearchTask is done, to set how many children tasks each SearchTask can generate after completion. Default is 2 that uses head-tail split, i.e., slice the shallowest unexplored choice as an independent SearchTask (head), and any remaining unexplored choices as another SearchTask (tail).
• DFS strategy has a single SearchTask at any given time.
• Random strategy can have many SearchTask. After a SearchTask completes, the next task is randomly chosen from the pending tasks list.
• A* strategy can have many SearchTask as well. After a SearchTask completes, the next task is chosen by preferring the one that is shallowest, by using search depth as the cost function.