Add preset-aware Hydra configuration UI

[suhanpark]

Summary

Expose GTSFM's Hydra configuration safely through Studio, with controls that follow the selected reconstruction and Gaussian-splatting presets.

Changes

• discover configuration schemas dynamically across 37 root presets without importing PyTorch during workspace startup
• expose 35 safe Gaussian-splatting parameters and preset-specific reconstruction fields
• load the schema lazily in the background so the main UI renders immediately
• hide internal paths and secret-bearing settings
• validate typed overrides on both the client and server before launching GTSFM
• add JSON configuration import with a generated template, nested or dotted keys, API aliases, and clear validation feedback
• pass validated reconstruction and Gaussian overrides through to the runner
• organize these controls in a collapsible Advanced Configuration section

Why

The command line already supports powerful Hydra presets and overrides, but users had to know their exact names and types. This adds a guided UI while preserving Hydra as the source of truth.

Validation

• 48 focused runtime and shutdown tests passed
• React/TypeScript production build passed
• Black, isort, and flake8 passed on changed Python files
• git diff --check passed

Stack

Depends on the dataset auto-detection draft and targets studio/dataset-autodetect to keep review scope focused.