Support aggregating multiple FST files

[TheZoq2]

I spoke with one of the Verilator maintainers at ORConf this weekend, and he asked if we could support "aggregated FST files" which as I understand it would significantly improve trace dumping performance in Verilator.

There are some more details about it here gtkwave/gtkwave#286

[ekiwi]

We could probably do that. Once GTKWave implements it, I might be motivated to take a look.
I do feel like one could also write a fairly performant multi-thread safe implementation of a FST write. However, with the original C code that might be a bit challenging.