ntp: use per-server chrony settings

[xiaotianlyu]

Description of changes:

Updates the shared NTP defaults to use named per-server chrony settings:

• configures 169.254.169.123 as a preferred server polled every 16 seconds;
• replaces 2.amazon.pool.ntp.org with the time.aws.com pool;
• enables chrony measurement, statistics, and tracking logs.

Adds a datastore migration for existing nodes. On upgrade, it converts the old settings.ntp.time-servers list and shared options into named servers. On rollback, it converts named servers back into a list and shared options that the old API can read.

The migration is required because Storewolf otherwise retains the old list while adding the new named keys. This was reproduced during upgrade as:

duplicate field time_servers

The migration is registered for the 1.64.0 -> 1.65.0 transition.

Testing done:

Local tests:

• cargo test -p ntp-time-servers: 6 passed
• cargo test -p migration-helpers: 19 passed
• cargo clippy -p ntp-time-servers --all-targets -- -D warnings: passed
• cargo fmt --manifest-path sources/Cargo.toml --all -- --check: passed
• git diff --check: passed

Filesystem migration test:

• migrated an old list datastore forward to named servers;
• migrated the named datastore backward to the old list form;
• verified the expected data and metadata in both directions.

Full image upgrade and rollback test:

• built a local core-kit and an aws-k8s-1.32 x86_64 1.65.0 test image containing all three related changes;
• generated a signed TUF repository and verified it contained migrate_v1.65.0_ntp-time-servers.lz4;
• launched a Bottlerocket 1.64.0 node with the old server list;
• upgraded the node from 1.64.0 to the test 1.65.0 image;
• rolled the node back to 1.64.0.

After upgrade:

• the migrator journal reported that the NTP migration completed successfully;

• apiclient get settings.ntp returned only the named link-local and amazon-pool entries, with no duplicate-field error;

• /etc/chrony.conf rendered:

  pool time.aws.com iburst
  server 169.254.169.123 prefer iburst minpoll 4 maxpoll 4
  driftfile /var/lib/chrony/drift
  makestep 1.0 3
  dumponexit
  dumpdir /var/lib/chrony
  logdir /var/log/chrony
  log measurements statistics tracking
  user chrony
  rtcsync
  

• changing logging to ["tracking"] rendered log tracking;

• changing logging to [] removed the log directive;

• chronyd was active with NRestarts=0;

• all three log files were written under /var/log/chrony with the expected
  owner and measure_t SELinux label.

After rollback:

• the node returned to Bottlerocket 1.64.0;
• the backward migration completed successfully;
• the old API read 169.254.169.123 and time.aws.com in the restored list;
• the old chrony template rendered both entries with the shared iburst option.

Related PRs:

• settings-sdk: feat: ntp: support per-server time source configuration bottlerocket-settings-sdk#146
• core-kit: chrony: render per-server time source configuration bottlerocket-core-kit#1030
• bottlerocket: ntp: use per-server chrony settings #4923

Testing remaining:

• after the core-kit release is available, rebuild the final image and launch a fresh
  instance;

• verify that chrony file logging is disabled by default;

• verify that /etc/chrony.conf contains no log directive;

• verify that logging can still be enabled through settings.ntp.logging.

Terms of contribution:

By submitting this pull request, I agree that this contribution is dual-licensed under the terms of both the Apache License, version 2.0, and the MIT license.

[ginglis13]

logging should be opt-in, not on by default

Suggested change

	# chrony log categories, rendered as the `log` line.
	logging = ["measurements", "statistics", "tracking"]

[ginglis13]

logging should be opt-in, not on by default

Suggested change

	# chrony log categories, rendered as the `log` line.
	logging = ["measurements", "statistics", "tracking"]

[xiaotianlyu]

Updated, thanks! Chrony file logging is now opt-in and is no longer enabled in the shared defaults.