Add speed benchmarks to CI

[abrookins]

Add performance benchmarks to CI to track model instantiation speed and prevent performance regressions.

Context

From discussion in #640, there's concern about the performance overhead of HashModel/JsonModel instantiation compared to plain Pydantic BaseModel. While some overhead is expected due to Redis OM's additional functionality (metaclass processing, PK generation, field validation), we should track this to:

1. Prevent performance regressions
2. Measure improvements from optimizations
3. Provide visibility into instantiation costs

Proposed Benchmarks

• Model instantiation timing (HashModel, JsonModel, EmbeddedJsonModel vs BaseModel)
• Bulk save operations
• Query performance (find, filter operations)

Implementation Options

• Use pytest-benchmark for benchmark tests
• Run benchmarks in CI and report results
• Consider using GitHub Actions for benchmark tracking (e.g., github-action-benchmark)

Reference

• Original performance analysis: Creating a model performance issue #640