With the work done under SERVER-58650 , a part of the splitVector code - used by the chunk splitter - has been mechanically ported into the autoSplitVector without much context.

As a precondition to SERVER-55028 that is aimed to improve such function, it is recommendable to further simplify/clarify the auto split vector code.

Planned refactoring:

• Extract the logic to extend the chunk bounds into an independent function
• Extract the minKey == maxKey check into an independent function
• Bring out of from the collection lock scope the code that doesn't require it
• Avoid the confusing part adding/removing minKey to the split vector, but rather simply keep the last split point reference initialized to the first chunk's key
• Better document each unintuitive step/statement
• Simplify code but preserve the original semantic