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

## 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