Implement adaptive should_run policies based on graph characteristics to optimize parallel execution

[Adrija-verse]

Problem
Currently, nx-parallel doesn't intelligently decide when parallelization is beneficial. This results in:

Wasted computational resources when parallel overhead exceeds any potential speedup
Suboptimal performance on small or dense graphs where sequential execution would be faster
Inconsistent user experience as speedup varies unpredictably across different graph types

Proposed Solution
Implement algorithm-specific should_run policies that dynamically determine whether to use parallelization based on graph characteristics and algorithm-specific heuristics.

[akshitasure12]

Hi @Adrija-verse, the should_run functionality has already been implemented in PR #123. That said, if you have ideas for improving this further, feel free to do so by raising a PR :)
Thanks!