Improve workload generation: schema auto-generation and nested document support

[janhoy]

Summary

The workload generator (osbenchmark/workload_generator/) has gaps around complex field
types and Solr-specific document structures. Two high-value improvements:

1. Schema auto-generation for complex field types

Current state: The generator handles basic field types but struggles with complex ones
(e.g. StrField with multiValued, TextField with custom analysers, PointField,
DateRangeField, geo fields, dynamic fields, copy fields).

Goal: Introspect a Solr schema and produce a valid, complete workload schema definition
including analyser chains, copy fields, and dynamic field patterns.

2. Nested / child document support

Current state: The generator assumes flat documents.

Goal: Support Solr's nested document model (parent/child via _nest_path_) so that
benchmarks can exercise {!parent} / {!child} queries and block-join indexing.

Acceptance criteria

• Generator correctly maps all standard Solr field types to OSBenchmark param types
• Dynamic fields and copy fields are reflected in generated workload schema
• Nested document corpora can be generated and indexed
• Existing generator tests updated / extended