improve efficiency of PDBEntry operations

Jalview's Sequence class holds lists of PDBEntry objects as a Vector. The list is populated when sequences with PDB crossrefs are imported, and can get extremely large, making operations on the list (searching by ID, updating existing entries with additional data) scale poorly.

Propose:
1. Move to a hash container (LinkedHashMap).
2. Review the need to have several PDBEntry objects with the same ID

Test data:
Stress test: retrieve AL123456 - this takes over an hour with Jalview 2.11.2.