push: cleaner prompts, rename --overwrite-version to -o, refactor RunE

[moshe-kabala]

What's new in leap push

Interactive flow

• Ask for the new version's name before the model file path. Matches how users mentally label a push ("what am I calling it → which file is it").
• Prompt label: "Push as new, or overwrite existing".
• When the user picks overwrite without --eval, a yellow NOTE points them at the UI's update-evaluate dialog (suppressed when --eval was passed).
• The auto-detect eval prompt is tighter: in-prompt hint instead of a separate header block, bold auto-detected and bold space in the multi-select instructions, and the instructions line now wraps onto its own line (template tweak — survey.MultiSelectQuestionTemplate overridden once at init). <right> / <left> select-all shortcuts removed from the hint.
• Plan output uses a single This will: verb across reset / update / no-actions branches; update branch always appends an Auto-detect added meta/viz, metric-direction, insight-config line so users see the full picture.

-o shortens --overwrite-version

leap push -o 6a16a0cf -e            # by id
leap push -o my-model -e            # by name (picker if ambiguous)

• --overwrite-version stays as a deprecated alias — existing scripts keep working but trigger cobra's one-line deprecation warning.

• ID lookups search across all versions (active or not); name lookups only consider active versions so the disambiguation picker matches what users see in the UI.

• The picker reuses the row format from the no-flag interactive flow.

• Non-TTY (CI) gracefully falls back to a candidate-ids error so scripts can re-run with an exact id.

• The resolved target is echoed before any destructive work:

  Overwriting model (id 6a16a0cf, matched "my-model")
  

--update accepts short aliases + implies --eval

leap push -o my-model -u viz                       # auto-implied --eval
leap push -o my-model -u metadata -u insights      # repeatable

short	(legacy still accepted)
metadata	update_metadata
metric	update_metric
insights	update_insights
visualization / viz	update_visualization

The previous "--update requires --eval" error is gone — passing -u now auto-sets --eval to true.

cmd/root_cmd/push.go refactor

The old 250-line RunE closure became a 40-line orchestration. Split into named helpers:

runPush(ctx, in)
├── validatePushInputs
├── newPushState                  (auth + workspace + project + analytics map)
├── resolveOverwriteTarget        (--overwrite flag OR interactive picker)
├── collectModelTypeAndName       (SelectModelType + InitMessage)
├── syncBranchSecretAndPython
├── resolveEvalPlan               → (batchSize, updateActions, runUpdateEvaluate)
├── pushCodeAndWaitForParsing
├── applyModelToVersion           (ImportModel OR OverrideModel)
├── sendSuccessEvent
└── triggerEvaluate               (RunUpdateEvaluateArtifact OR RunEvaluate)

Single s.fail(stage, err) helper tags analytics and emits the fail event — replaces the ~15 inline-duplicated tagging blocks the old RunE had.

Docs

New docs/push-examples.md with a quick tour and a cheat sheet.

Test plan

• go build ./... clean
• go vet ./... clean
• leap push --help renders examples + new -o flag
• Interactive leap push against a project with sessions + content — confirm name prompt before path
• Interactive leap push -e against a project with an overwrite-able version — confirm new "Push as new, or overwrite existing" prompt + the multi-select with bold "auto-detected" and bold "space"
• leap push -o <id> — confirm "Overwriting … (id …)" echo
• leap push -o <duplicate-name> -e — confirm picker opens
• leap push -o <name> -u viz — confirm runs without explicit -e
• leap push --overwrite-version <id> -e — confirm deprecation warning