[docs] Document local log TTL and active segment rolling

[wuchong]

Motivation

Local log retention behavior is being extended by #3748 and #3830. The user-facing documentation should describe the resulting configuration model and clearly distinguish remote log retention, local log retention, and active-segment rolling. It should also provide safe rollout guidance for clusters upgrading from 0.9 to 1.0.

Documentation scope

• Document table.log.local-ttl:
  • it controls retention of local log segments;
  • a local segment is eligible for TTL cleanup only after it has been copied to remote storage;
  • document its default value and validation constraints, including its relationship with table.log.ttl;
  • reconcile with the initial documentation changes already included in [server] Support independent local log TTL #3778.
• Document log.retention.roll-active-segment.enabled:
  • it is a server configuration and is disabled by default;
  • when enabled, a non-empty active segment may be rolled after the table log TTL expires and the high watermark reaches the log end offset;
  • explain that rolling makes the segment eligible for remote upload and subsequent local cleanup.
• Expand the TTL page to distinguish:
  • table.log.ttl for table/remote log retention;
  • table.log.local-ttl for copied local segments;
  • active-segment rolling controlled by log.retention.roll-active-segment.enabled.
• Update or regenerate the relevant table and server configuration reference pages.
• Update the tiered-storage documentation and add cross-links to the TTL page.
• Update the 1.0 upgrade notes for clusters upgrading from 0.9:
  • keep log.retention.roll-active-segment.enabled disabled during the upgrade; this is also the default value;
  • after all cluster components have been upgraded to 1.0 and the upgrade is complete, enable the option through a dynamic cluster configuration update;
  • explain that enabling the option allows expired active local log segments to be rolled and subsequently cleaned up automatically.

Related issues and pull requests

• Related to [Feature] Support independent TTL configuration for local log segments #3748
• Related to [server] Active log segment is not rolled after table.log.ttl expires #3830
• Related PR: [server] Support independent local log TTL #3778
• Related PR: [server] Roll expired active segments for remote retention #3831

Acceptance criteria

• Both configuration keys are searchable in the published documentation.
• Their defaults, constraints, scope, and interaction are documented.
• The TTL page explains the local-segment lifecycle from active segment, to rolled/uploaded segment, to local cleanup.
• The 1.0 upgrade notes document the staged 0.9-to-1.0 rollout: keep log.retention.roll-active-segment.enabled disabled during the upgrade, then enable it with a dynamic cluster configuration update after the upgrade is complete.
• Existing documentation changes in [server] Support independent local log TTL #3778 are retained and reconciled instead of duplicated.
• The TTL, configuration-reference, tiered-storage, and upgrade-note pages link to each other where appropriate.

[wuchong]

cc @gyang94