CASSANDRA-20446: Removed parsing of string for MBean creation

[omniCoder77]

JIRA Ticket

Issue:
For MBean creation, cassandra uses javax.management.ObjectName(String name) constructor. Due to this, string has to be parsed to extract properties.

My Proposed Solution:
Use javax.management.ObjectName(String name, Hashtable<String, String> table) constructor. Using this we skip parsing.

Validation
All the CassandraMetricsRegistryTest.java pass

jmap before patch

jmap after patch

Before patch

1. ObjectName$Property: 56,206 instances (~1.35 MB)
2. ObjectName$Property[]: 29,383 instances (~0.93 MB)
3. String: 107,906 instances

After patch

1. ObjectName$Property: 18,763 instances (~0.45 MB)
2. ObjectName$Property[]: 19,023 instances (~0.46 MB)
3. String: 96,639 instances
4. ObjectName: 9,512 instances (~0.30 MB)

This represents ~66% reduction in ObjectName$Property instances and a decrease in overall JMX's heap usage, and no changes to the externally visible JMX names or behavior.

[omniCoder77]

@netudima @smiklosovic Requesting your reviews when convenient, this PR implements the JMX ObjectName deduplication suggested in CASSANDRA-20446.