[Flink] Fix REMOTE Session JobGraph submission for Flink 1.x/2.x SQL and uploaded JAR jobs

[shangeyao]

Summary

Fixes REMOTE Session JobGraph submission for Flink 1.17–1.20 and 2.x SQL / uploaded JAR applications (Closes #4506).

This PR builds on #4500, which already landed the Flink 2.x REMOTE JobGraph baseline on dev. This change adds the remaining cross-version fixes discovered during local e2e validation.

SQL

• Normalize storage/decoding with DeflaterUtils.toPlainText / compressForStorage
• Harden FlinkSql.checkChange for null/compressed input
• Sync remote cluster versionId on update

Uploaded JAR

• Skip distHome for upload resources; fix team-scoped upload paths
• Resolve shaded jar fallback for upload jobs
• Add Flink runtime classpath + parent-first classloading for JobGraph build (including Flink 1.20 flink-dist)

Client / packer / shims (1.x + 2.x)

• Read main class from JSON appConf (SubmitRequest.appMain)
• Fix ShadedBuildResponse JSON mapping
• SQL fat-jar snakeyaml handling (sqlclient shade, MavenTool, build pipeline lib copy)
• Flink 2.x shims-base upload in EnvInitializer
• FlinkShimsProxy parent-first for org.yaml
• JDK 17 SecurityManager tolerance; REST fallback via HttpClient

Relationship to #4500

Layer	PR	Responsibility
Baseline	#4500 (merged)	Flink 2.x REMOTE JobGraph submit path, classloader/ServiceLoader fixes
This PR	#4507	SQL storage, upload JAR build/submit, snakeyaml/classpath, 1.20 flink-dist, console fixes

Test plan

• Flink 1.20.5 REMOTE Session: SQL build + start → RUNNING via JobGraph (no REST fallback)
• Flink 1.20.5 REMOTE Session: uploaded JAR (Flink 1.20-compatible examples jar) → RUNNING via JobGraph
• Flink 2.x REMOTE Session: SQL + JAR regression verified in local e2e (on top of [Flink] Make Flink 2.x job submission work on REMOTE mode #4500)
• CI

Note: uploaded JARs must match the target Flink version.

[sonarqubecloud]

Quality Gate passed

Issues
9 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud