Malloc tweak

Improve OpenZGY read performance when the application requests just a single brick or part of a brick at a time, from multiple threads, and is run in a high-performance (>= 48 core machine) on Azure (i.e. not Google) with co-located data storage, and the application isn't doing other work than reading data. For Petrel this is mainly when using "Prefetch to cache" but there may be other workflows where the tweak is noticeable.

The tweak is automatically skipped if reading larger areas that make the consolidate-bricks logic kick in. Apart from that the tweak will be active. But except for the case described above it probably won't have any measurable effect. Because there will be other factors that slow down the throughput a lot more.

More description will be added as comments. The text in the description box is only allowed 4000 characters.