feature: Multicast integration

[Lore0599]

Multicast integration

This PR introduces multicast capabilities to the snitch cluster.

🆕 ADDED:

• 🧩 CFG: The snitch configuration file can now accept some new parameters:

  • dma_user_width: Specifies the width available to encode the multicast mask on the wide interconnect.
  • enable_multicast: Boolean flag to enable a multicast-capable crossbar on the wide interconnect.
  • atomic_id_width: Defines the number of bits used in the user field to support atomic operations. This is required to distinguish between hart ID and the multicast mask.

• ⚙️ HW:

  • axi_mcast_xbar: The AXI crossbar has been extended with multicast support to enable multicast communication over the wide interconnect.
  • CSR_MCAST: A new CSR has been added to the Snitch core to allow software to encode and configure the multicast mask.
  • snitch_lsu: The LSU has been extended to forward the multicast mask correctly through the transaction's user field.

• 💻 SW:

  • The snRuntime has been updated to support multicast in synchronization mechanisms, such as global barriers.
  • The runtime has also been modified to address race conditions in multi-cluster systems. Specifically:
    • The offload mechanism now requires waking up core 0 first.
    • Core 0 is responsible for clearing the BSS memory region.
    • Only afterwards are the other clusters woken up by core 0.

• CI

  • Convert repo names to lowercase for use in Docker tag name