redesign caches and clear/disable during tests

[Jutho]

Currently we have the following cache variables (I hope I am not missing any):

TensorKit.GLOBAL_FUSIONBLOCKSTRUCTURE_CACHE
TensorKit.treetransposercache
TensorKit.treebraidercache
TensorKit.treepermutercache
TensorKit.transposecache
TensorKit.braidcache

As well as the control flags

TensorKit.usebraidcache_abelian
TensorKit.usebraidcache_nonabelian
TensorKit.usetransposecache
TensorKit.usetreetransposercache
TensorKit.usetreebraidercache
TensorKit.usetreepermutercache

and

abstract type CacheStyle end
struct NoCache <: CacheStyle end
struct TaskLocalCache{D<:AbstractDict} <: CacheStyle end
struct GlobalLRUCache <: CacheStyle end

function CacheStyle(I::Type{<:Sector})
    return GlobalLRUCache()
end

fusionblockstructure(W::HomSpace) = fusionblockstructure(W, CacheStyle(sectortype(W)))

for the fusion blockstructure caches.

We should probably make this a bit more uniform. Importantly, we should disable or empty caches while running the tests, as they in the end collect close to a gigabyte of cached data, with probably only negative effects on the test performance.

[lkdvos]

I would say that the most important parts would be the ability to visualize and monitor cache usage, as well as easily clearing it. Also, I think we should be a little more conservative with the default values for the maximal sizes.

As a side note, we also have some caches for Fsymbol and Rsymbol, but only for some sector types (SUNRepresentations.jl mostly). Do we also want to include these in this discussion?

[lkdvos]

Closed since #244