docs: sync content from legacy Docusaurus repo

[divyasinghds]

Summary

Port recent content updates from the legacy tracebloc/documentation (Docusaurus) repo into the Mintlify docs site. Covers 12 commits that landed in the legacy repo after the Mintlify split.

• create-use-case/prepare-dataset: <workspace> placeholder, host-PVC ingestion (replaces COPY-into-image), Docker Hub setup section, ingestor-job.yaml cleanup, multi-arch buildx (linux/amd64,linux/arm64)
• create-use-case/prerequisites: add Tabular Regression / Time Series Forecasting / Time to Event Prediction; fix label CSV columns and example links; add prominent "filename omits extension" Note; mark Keypoint Detection / Semantic Segmentation as "template coming soon"
• environment-setup (configuration, eks-client-deployment-guide, setup-guide): align with unified tracebloc/client Helm chart 1.3.x; installer URL → i.sh / i.ps1; corrected PVC keys (pvc.{data,logs,mysql}); PSA labels + dedicated tracebloc-node-agents namespace; --reset-then-reuse-values upgrade flow
• join-use-case (hyperparameters, start-training): migrate API examples to snake_case (upload_model, link_model_dataset, learning_rate, loss_function, early_stop_callback, …) for tracebloc_package 0.7.0; rename trainingObject → training; correct terminate-on-NaN wording; pip optional-extras syntax

Skipped: azure-deployment-guide and getting-started/quick-setup — neither has a target file in the Mintlify repo.

Test plan

• mint dev renders all updated pages without MDX errors
• mint broken-links is clean (cross-page links repointed to Mintlify-style /section/page paths)
• Spot-check that Mintlify <Note> / <Warning> components render where the legacy used :::note / :::caution
• Verify code snippets (Helm install, buildx, ingestor-job.yaml) match the current tracebloc/client and tracebloc/data-ingestors repos

🤖 Generated with Claude Code

---

Note

Low Risk
Documentation-only changes, but they alter operational commands and config snippets (Helm/installer URLs, storage, and ingestion paths), so mistakes could cause user deployment failures.

Overview
Updates docs to align dataset ingestion with the client-host PVC workflow: datasets are placed in ~/.tracebloc/<workspace>/data and mounted into the ingestor pod at /data/shared (no more COPYing data into images), with clarified LABEL_FILE/SRC_PATH, <workspace> namespace usage, and multi-arch docker buildx (linux/amd64,linux/arm64) guidance plus a new Docker Hub setup section.

Expands and corrects prerequisites for supported tasks by adding tabular Regression, Time Series Forecasting, and Time to Event Prediction, and clarifies label CSV conventions (notably filenames omit extensions and object detection uses image_label).

Refreshes environment setup docs for the unified tracebloc/client Helm chart and newer installer endpoints (i.sh/i.ps1), including updated values structure (pvc.{data,logs,mysql}), platform-specific storage snippets, --reset-then-reuse-values upgrade flow, and added guidance on auto-upgrade, NetworkPolicy hardening, PSA labels, and the privileged tracebloc-node-agents namespace for the resource monitor.

^{Reviewed by Cursor Bugbot for commit b051bb4. Bugbot is set up for automated code reviews on this repo. Configure here.}