Move float bulk-write into IndexWriter to enforce endianness

[abernardi597]

In most places, the byte order is BIG_ENDIAN due to:

• DataInput/DataOutput method contracts
• Explicit calls to ByteBuffer.order(ByteOrder.BIG_ENDIAN)

The exception is the bulk-float write call in ArrayVectorProvider, which will default to the native byte-order.

While changing the byte order there to BIG_ENDIAN would fix this, it still does not allow consumers of this library to override this behavior (where they may wish to read LITTLE_ENDIAN in their implementation of RandomAccessReader due to the static initialization of the VectorTypeSupport.

So this PR adds a bulk writeFloats to IndexWriter and uses that interface over DataOutput where necessary.
I also took the liberty to add a missing .order(BIG_ENDIAN) call in SimpleMappedReader.

Unrelated changes include:

• Adding jdtls LSP files to .gitignore
• Fixing a pre-existing compilation error in ProductQuantization related to type inference of Runnable rather than Callable

Closes #576.