deps(ipfs): bump kubo to v0.41.0 and fix removed/no-op config

[odesenfans]

Summary

Bumps the IPFS image from kubo:v0.37.0 to kubo:v0.41.0 across all
deployment compose files and fixes two configuration items that became
broken in the intervening releases.

1. Reprovider.* config keys REMOVED in 0.41

Kubo 0.38 merged the Provider and Reprovider sections into a single
Provide section. The v0.41 docs/config.md
explicitly marks Reprovider.Strategy as REMOVED (and
Reprovider.Interval → use Provide.DHT.Interval).

Our init script set this value on every container start
(ipfs config Reprovider.Strategy 'pinned'), so --migrate does not
help: the line errors on each boot. Switched to Provide.Strategy.

2. --enable-pubsub-experiment is now a no-op (silent breakage)

Per v0.41 source
(cmd/ipfs/kubo/daemon.go):

cmds.BoolOption(enablePubSubKwd, "DEPRECATED CLI flag. Use Pubsub.Enabled config instead.")
...
if psSet {
    log.Error("The --enable-pubsub-experiment flag is deprecated. Use Pubsub.Enabled config option instead.")
} else {
    pubsub = cfg.Pubsub.Enabled.WithDefault(false)
}

The flag value is no longer wired up, and Pubsub.Enabled defaults to
false. Since pyaleph relies on IPFS pubsub for peer liveness
(src/aleph/services/ipfs/pubsub.py, services/peers/publish.py,
services/peers/monitor.py), leaving this unchanged would silently
disable pubsub.

Fix: removed the dead flag from the daemon command and added
ipfs config Pubsub.Enabled --json 'true' to the init script.

Preserved

• --migrate is kept so the on-disk repo migrates from v17 to v18 on
  first boot under 0.38+.

Operator upgrade notes

• On nodes whose /data/ipfs volume was previously owned by root
  (e.g. an older deployment that ran the container as root), the new
  image runs as ipfs (UID 1000) and the entrypoint's auto-chown can
  be slow on large volumes. Operators may want to run a manual
  chown -R 1000:100 /var/lib/docker/volumes/<...>/_data ahead of the
  bump. This is unchanged behaviour vs. 0.37 (the kubo Dockerfile has
  set ipfs:users ownership for years); only volumes with mismatched
  history are affected.

Test plan

• Deploy on a staging node previously running kubo:v0.37.0,
  confirm: daemon starts, ipfs config Pubsub.Enabled returns
  true, ipfs config Provide.Strategy returns pinned, no
  Reprovider.Strategy errors in 001-update-ipfs-config.sh logs.
• Verify pyaleph peer liveness messages keep flowing on the IPFS
  pubsub topic after upgrade.
• Confirm repo migration v17 → v18 runs once on first boot.

[foxpatch-aleph]

Correctly bumps kubo from v0.37.0 to v0.41.0 across all 4 compose files. Replaces the removed Reprovider.Strategy config key with Provide.Strategy, removes the no-op --enable-pubsub-experiment daemon flag (sets Pubsub.Enabled via config instead), and preserves all other settings. All changes are verified against kubo v0.41.0 source and config docs. Clean, well-researched, and focused PR.

deployment/scripts/001-update-ipfs-config.sh (line 21): Verified against kubo v0.41.0 source: Pubsub.Enabled is type flag (boolean, default false), and --json 'true' correctly sets a JSON boolean. The --enable-pubsub-experiment daemon flag was confirmed to be a no-op (only logs deprecation error, value not wired up) at cmd/ipfs/kubo/daemon.go:424-428.

deployment/scripts/001-update-ipfs-config.sh (line 17): Verified against kubo v0.41.0 config.md: Reprovider.Strategy is REMOVED and replaced by Provide.Strategy. Value 'pinned' is a documented valid strategy.

deployment/docker-build/docker-compose.yml (line 55): Pre-existing (not introduced by this PR): bind mount ./001-update-ipfs-config.sh does not exist at deployment/docker-build/001-update-ipfs-config.sh. The script lives at deployment/scripts/001-update-ipfs-config.sh.

deployment/docker-build/dev/docker-compose.yml (line 125): Pre-existing (not introduced by this PR): same bind mount issue as the other compose files.

deployment/samples/docker-compose/docker-compose.yml (line 113): Pre-existing (not introduced by this PR): same bind mount issue as the other compose files.

deployment/samples/docker-monitoring/docker-compose.yml (line 115): Pre-existing (not introduced by this PR): same bind mount issue as the other compose files.