MPI_DP: Improve scalability

Multiple changes are contained in this commit:

- Adoption of BSD Queues instead of custom made linked-lists.
- Used of linked-lists in often-insertion/no-random-access scenarios.
- Removed unneeded dynamic allocated variables.
- Fixed minor memory leak on StreamWSR destruction.
- Finer grain lock/mutex usage.
- Added multiple comments and diagram.
- Renamed most of the data structures for better consistency.
- Rearranged function order in the file to better follow its execution.