Should we use the unpooled allocator in the transport client? #34330
Description
The default (Netty) allocator for Elasticsearch is pooled (since #30684 only if the heap size is > 1GB). This makes sense on the server but the transport client also uses the pooled (Netty) allocator. Contrary to the server it could make sense to switch the client to the unpooled allocator (trading memory footprint for more frequent GCs). To make this decision we need to run benchmarks first to gather more data.