HBASE-30385 Fix duplicate system region move scheduling

[Huginn-kio]

HBASE-30385

Summary

AssignmentManager.checkIfShouldMoveSystemRegionAsync() can schedule the same
system-region move more than once when multiple excluded RegionServers are
processed. This can cause a subsequent move submission to fail because the
region is already in transition. See HBASE-30385.

Root cause

On Apache master, plans is collected before the loop over excluded servers,
but the loop that submits the collected plans is inside that outer loop
(hbase-server/src/main/java/org/apache/hadoop/hbase/master/assignment/AssignmentManager.java,
around lines 741-745). Plans collected from an earlier server are therefore
submitted again for each later excluded server. The log also reports the outer
loop variable instead of the plan source.

Fix

• Submit the collected RegionPlans after processing all excluded servers.
• Log plan.getSource() so the reported source matches the plan being moved.

Testing

• Failing-before-fix evidence: The duplicate scheduling is demonstrated by
  the existing control flow and the reported in-transition exception; no
  standalone regression test was added.
• Passing-after-fix evidence: git diff --check passed.
• Neighbor regression slice: Not run; no unit test was added for this
  small control-flow correction.
• Static analysis: Targeted checkstyle completed with no violations on the
  modified lines. The full file still reports three pre-existing violations
  outside this diff.

Compatibility

• Public API impact: none
• Config impact: none
• Persisted format impact: none
• Wire format impact: none