Fix JVM fingerprinting by adding tags (#8175)

# Problem

Tags could decide how a target is compiled (rsc, zinc), but we are not factoring them into the cache key, causing RscCompile task to hit the cache that should not be hit.

# Solution

Explicitly add tags to target's fingerprint.

src/python/pants/backend/jvm/subsystems/dependency_context.py

@@ -10,6 +10,7 @@
 from pants.backend.jvm.targets.javac_plugin import JavacPlugin
 from pants.backend.jvm.targets.scalac_plugin import ScalacPlugin
 from pants.base.fingerprint_strategy import FingerprintStrategy
+from pants.base.hash_utils import stable_json_sha1
 from pants.build_graph.dependency_context import DependencyContext as DependencyContextBase
 from pants.build_graph.resources import Resources
 from pants.build_graph.target_scopes import Scopes
@@ -86,6 +87,8 @@ def compute_fingerprint(self, target):
 
     hasher = hashlib.sha1()
     hasher.update(target.payload.fingerprint().encode())
+    # Adding tags into cache key because it may decide which workflow applies to the target.
+    hasher.update(stable_json_sha1(target.tags).encode())
     if isinstance(target, JarLibrary):
       # NB: Collects only the jars for the current jar_library, and hashes them to ensure that both
       # the resolved coordinates, and the requested coordinates are used. This ensures that if a