feat(vertex): add default dimensions parameter to VertexAIEmbeddings

[saeedseyfi]

Summary

Add optional dimensions constructor argument to VertexAIEmbeddings class. This allows users to set a default output dimensionality that will be used for all embedding requests unless explicitly overridden.

Changes

• Add dimensions field to VertexAIEmbeddings with default value None
• Update embed() method to use constructor default when dimensions not specified in method call
• Explicit dimensions in method calls override the constructor default
• Full backward compatibility maintained - existing code continues to work unchanged

Usage Example

# Set default dimensions in constructor
embeddings = VertexAIEmbeddings(model="text-embedding-004", dimensions=128)

# Uses 128 dimensions (from constructor)
embeddings.embed(["hello", "world"])

# Overrides with 256 dimensions
embeddings.embed(["hello", "world"], dimensions=256)

# Backward compatible - no dimensions specified
embeddings = VertexAIEmbeddings(model="text-embedding-004")
embeddings.embed(["hello", "world"])  # Uses model's default

Test Plan

• Added 5 new unit tests covering all scenarios
• Test that constructor dimensions are used when not specified in embed()
• Test that explicit dimensions override constructor default
• Test backward compatibility when no default dimensions specified
• Test default dimensions work with embed_documents() and embed_query()
• All 10 unit tests pass
• Code quality checks pass (lint, format, mypy)

[saeedseyfi]

Hey @lkuligin, thanks for your approval, would you merge as well?