use MemoryPool instead of allocator, fix examples

[zxcv05]

MemoryPool is more performant and efficient than using a generic allocator, especially for lots of create()s and destroy()s of only a single type (Node(K, V, D) in this case)

also refactored part of the BPlusTree struct for easier readability and nicer qualities like BPlusTree(K, V, D).Node instead of having detached BPlusTree(K, V, D) and Node(K, V, D) (they are intrinsically linked)

[dayvster]

Sorry, life got in the way and github is not really all that great at notifying people on PRs.

This looks good, I'll do some testing/playing around with this today, run some benchmarks and if it's all positive merge it into main.

Appreciate it :)

[dayvster]

@zxcv05 had some time to play around with this, I have some thoughts.

I like where you're going with this. MemoryPool(Node) is a solid upgrade, the new init(allocator) style is cleaner, DebugAllocator in the examples is much better, and nesting Node inside BPlusTree feels nicer. Good stuff here.

That said, I can't merge it yet

• removeNode and the rebalancing logic have some inconsistent parts (especially around freeNode calls)
• createNode doesn't always initialize fields properly

What needs fixing/attention:

• run zig fmt some formatting is inconsistent
• Clean up inconsistent freeNode / pool.destroy usage
• Finish (or properly stub) the remove path