docs: flesh out templates.mdx and rewrite faqs.mdx with real user questions

[saadqbal]

Two docs files are weaker than the rest of the site:

templates.mdx — currently hollow

The file is 7 lines: just a link to the website's explore section. There's no on-page list of available templates, no descriptions, no direct links to the runnable data-ingestors templates (image_classification, object_detection, text_classification, tabular_classification, tabular_regression, time_series_forecasting, time_to_event_prediction).

faqs.mdx — feature-checklist tone

The current FAQ reads like an internal checklist ("System Requirements", "Training Questions", "Common Issues") rather than answers to questions users actually ask. Real user fears around tracebloc are about data privacy ("Can the data scientist see my raw data?"), IP ("Where do model weights end up?"), infrastructure ("Do I need a GPU?", "Air-gapped support?"), and operations ("What if training fails?"). None of those are addressed directly.

Acceptance criteria

• templates.mdx lists all 7 working templates with direct GitHub links + brief usage flow.
• templates.mdx notes that keypoint detection and semantic segmentation templates are pending (matching the "coming soon" footnotes in prerequisites.mdx).
• faqs.mdx leads with privacy questions (the most common fear), grouped into Privacy & Security / Infrastructure / Training & Models / Cost & Support.
• Each FAQ entry has a real answer (not "see other doc"), with cross-links where appropriate.
• Tone matches the rest of the docs (active voice, second person, sentence-case headings).
• Use the existing terminology from key-terms.mdx: "data owner", "data scientist", "client", "use case" — not "vendor".