Fix grouped MAX for non-positive floating-point values

[JackieTien97]

Description

Fix table-model grouped MAX returning NULL when every FLOAT or DOUBLE
value in a group is zero or negative.

GroupedMaxAccumulator initialized floating-point states with Float.MIN_VALUE
and Double.MIN_VALUE. In Java, those constants are the smallest positive
values, not the most negative values. As a result, non-positive inputs never
initialized the group state, and the final result was emitted as NULL.

MAX_BY was unaffected because its implementation explicitly handles the first
input value.

Changes

• Initialize grouped floating-point MAX states with negative infinity.
• Add unit coverage for zero, negative, and negative-infinity values for both
  raw and intermediate inputs.
• Verify that groups with no input still return NULL.
• Add a table-model SQL regression test covering the non-stream grouped
  aggregation path.

User impact

Grouped MAX(FLOAT) and MAX(DOUBLE) now return the correct value for groups
whose maximum is non-positive, including all-zero groups.

Validation

mvn clean test -pl iotdb-core/calc-commons \
  -Dtest=GroupedMaxAccumulatorTest

mvn clean verify -Drat.skip=true -DskipUTs \
  -Dit.test=IoTDBTableAggregationNonStreamIT#maxNonPositiveFloatingPointTest \
  -DfailIfNoTests=false -Dfailsafe.failIfNoSpecifiedTests=false \
  -pl integration-test -am -PTableSimpleIT -P with-integration-tests

The targeted integration test passed and the 37-module clean reactor build
completed successfully.

[CritasWang]

I think this PR needs to be updated to 2.0.11