Support Tile Clusters (tree-like hierarchies)

[jerryz123]

This work substantially refactors the existing Tile/TilePRCIDomain/HasTiles/InstantiatesTiles code into a system that support recursive clusters of tiles.

This PR defines an Element as an abstract node in the hierarchy tree. Elements can be BaseTiles or Clusters, where BaseTiles are leaf nodes in the hierarchy, clusters are non-root nodes, and the BaseSubsystem is the root node.
BaseTile and Cluster will support deduplication.

Notably Clusters support recursive instantiation, a Cluster can contain Tiles or internal Clusters. The Cluster bus topology should be capable of both supporting hierarchical coherence, and the BaseSubsystems topology handle managing all coherence, depending on configuration variables.

This PR still requires more time. Some of the interrupt handling is not satisfying, and some decisions over clock/reset crossing ought to be addressed.

Type of change: other enhancement

Impact: API modification

Development Phase: implementation

Release Notes

[sequencer]

will this PR block the 3.6 bump?

[jerryz123]

No. I don't think this PR should block anything, we should delay features while more critical things are going in.
(I certainly want to see more thorough testing before this goes in).

[sequencer]

the semantic of hartid and tileid is different in MT cores(although there are no opensource MT core for now). This PR seems to remove the hartid, is this a desidered refactor?

[sequencer]

No need in 3.6

[jerryz123]

the semantic of hartid and tileid is different in MT cores(although there are no opensource MT core for now). This PR seems to remove the hartid, is this a desidered refactor?

tileId is static, and should only be referenced by generator (its the old "staticIdForMetadataUseOnly").
hartId should be set by IO to allow dedup.

Future work can allow a tile to support multiple hartid.

[sequencer]

Great!

[jerryz123]

Replaced by #3490