Add Spark connector for TsFile table model

[DaZuiZui]

Motivation

Closes #843.

Changes

• Add the java/spark-tsfile Maven module and register it under the Java build.
• Implement a Spark 3.x DataSource V2 connector with the short name tsfile for TsFile table model batch reads and writes.
• Reuse existing TsFile Java APIs for table metadata, query execution, table schema registration, tablet writes, and TsFile writing.
• Support table selection, single-table inference, multi-file schema compatibility checks, sparse FIELD nulls, column pruning, and basic time/TAG predicate pushdown.
• Support DataFrame writes to table-model TsFile directories, with one part-*.tsfile per non-empty Spark task partition.
• Harden append writes so output file names include the Spark query id, commit refuses to overwrite existing files, and abort only removes temporary files.
• Document the initial local-file-only scope and current follow-up areas.

Tests

• JAVA_HOME=$(/usr/libexec/java_home -v 21) ./mvnw -P with-java -pl java/spark-tsfile -am test
• Spark connector targeted read/write smoke tests for round trip, append twice, SQL temporary view, external schema validation, and partitioned writes.

Notes

The initial connector intentionally supports local TsFile paths only. Non-local Hadoop filesystems, schema merging, broader predicate pushdown, streaming semantics, expanded categories/types, and ResourceBundle-backed connector exception messages should be handled in follow-up changes.