Cardano Updates

- Add `profiles: ["devnet"]` to all proto-devnet services
- Proto-devnet now requires `--profile devnet` to start
- ImmDB remains `--profile immdb`
- Update README with new usage instructions
- Update IP addresses in README for immdb profile

Usage:
  docker compose --profile devnet up      # Proto-devnet
  docker compose --profile immdb up       # ImmDB mock

Signed-off-by: Chris Gianelloni <[email protected]>

Documented regression for ledger apply/reapply

- Separate immdb profile IPs to avoid conflicts with proto-devnet:
  - upstream: 172.28.0.10 -> 172.28.0.110
  - node0: 172.28.0.20 -> 172.28.0.120
  - downstream: 172.28.0.30 -> 172.28.0.130
- Fix genesis file paths: use /data directly instead of /data/genesis
  to match config.json relative path expectations
- Both profiles now work correctly together or separately

Signed-off-by: Chris Gianelloni <[email protected]>

- cardano-node from leios-prototype branch
- tx-generator from stable cardano-node 10.5.4 (leios-prototype has API incompatibility)
- cardano-cli from blink labs pre-built image

The tx-generator on leios-prototype fails to compile due to
ouroboros-network API drift, but has no Leios-specific changes.

Signed-off-by: Chris Gianelloni <[email protected]>

The prot-devnet setup was creating too many streams. Loki treats each
unique 'label' (a log namespace) as a stream AFAIK

Rewrite mempool-sim-web core to use:
- Global tx registry with per-tx BitSet bitmaps (1 bit/node)
- Lightweight nodes (counters only, no tx storage)
- Same mini-protocol: OfferTx → RequestTx → SendTx
- Same link delay model (latency + size/bandwidth)
- Same adversary model (intercept, create variant, frontrun delay)
- Endorser block support (--eb flag) with tx cache modes
- 1000 nodes / 18M events in ~9s / 130MB heap

New files: bitmap.ts, tx-registry.ts
Removed: mempool.ts, peer-manager.ts (functionality merged into bitmaps)

Mathematical model of mempool diffusion

- Fix timestamp parsing to handle nanosecond precision (truncate to microseconds)
- Fix log format detection: use 'ns' field for namespace detection (cardano-node format)
- Add type safety checks for msg/data fields that may be strings
- Parse block hash and slot from 'data.newtip' field (format: "hash@slot")
- Detect BlockFetch.CompletedBlockFetch and immdb-server MsgBlock events
- Fix schedule cleanup: remove stale base-schedule.json and schedule.json on init
- Reduce default LEIOS_RELEASE_TIME from 128.9s to 10s for faster testing

Signed-off-by: Chris Gianelloni <[email protected]>

Was incorrectly set to 90KB (RB size). EBs are the throughput
expansion mechanism — 10 MB is the CIP default.

Leios mempool now: 2 × (90KB + 10MB) ≈ 20.18 MB

Praos: mempool = 2 × RB_size (unchanged)
Leios: mempool = 2 × (RB_size + EB_size)

Removes the crude leiosMempoolMultiplier in favor of principled
sizing that automatically scales with EB size.

- Remove obsolete version attribute from docker-compose.yaml
- Rename compose.yaml to k8s-pod.yaml to avoid conflict with
  docker compose (was a Kubernetes Pod spec, not a compose file)

Signed-off-by: Chris Gianelloni <[email protected]>