[core] Support aggregation for partial-update sequence groups

[JingsongLi]

Purpose

Allow Partial Update tables to configure one aggregate function for every value field protected by a sequence group, using a sequence-centered option:

'fields.seq.sequence-group' = 'a,b,c',
'fields.seq.sequence-group.aggregate-function' = 'sum'

Per-field aggregate functions still take precedence, followed by the sequence-group aggregate function and then the default aggregate function. Ordering and primary-key fields are never aggregated.

This change also validates orphan sequence-group aggregate options, preserves the option during column renames, and documents the new syntax.

Tests

• mvn -pl paimon-core -am -DfailIfNoTests=false -DwildcardSuites=none -Dtest='CoreOptionsTest#testSequenceGroupAggregateFunction+testSequenceGroupAggregateFunctionWithoutGroup,SchemaManagerTest#testRenameColumnInSequenceGroupAggregateFunction' test
• The new merge behavior test and the broader related test classes passed on the implementation baseline. On current master, local execution of both the new test and the existing PartialUpdateMergeFunctionTest#testSequenceGroup is blocked by the same JDK 8 CodeGenLoader / URLClassPath NPE before merge assertions run.