## Summary

- `ensure_installed` walked up to the nearest `package.json`, which in a
monorepo subpackage stops at the subpackage. The install state file only
exists at the workspace root, so every `aube run` / `exec` / `start`
from a subpackage reported `install state not found` and spuriously
re-ran install even when the root tree was fresh.
- Prefer `find_workspace_root` as the freshness anchor, fall back to
`find_project_root` for non-workspace projects, and finally to the cwd.
- Added a regression test: `aube install` at the workspace root, then
`aube start` from `packages/app` must not print `Auto-installing`.

## Why

User report (bug #3): `aube start` in a workspace package re-ran install
on every invocation after a root `aube install`. The freshness check was
reading the wrong `node_modules/.aube-state` path.

## Test plan

- [x] `cargo clippy --all-targets -- -D warnings`
- [x] `cargo test --release -p aube` — 278 unit + 4 e2e pass
- [x] `mise run test:bats test/run.bats` — 17 tests pass, including the
new regression

<!-- CURSOR_SUMMARY -->
---

> [!NOTE]
> **Medium Risk**
> Changes the directory used to determine dependency freshness for
auto-install, which can affect when installs run in monorepos vs
single-package projects. Risk is limited by simple fallback logic and
added end-to-end regression coverage.
> 
> **Overview**
> Fixes auto-install staleness checks in workspaces by anchoring
`ensure_installed` at the *workspace root* (falling back to the nearest
`package.json`, then the current directory) so subpackages reuse the
root `.aube-state` instead of repeatedly reporting missing state and
re-installing.
> 
> Adds a Bats regression test that runs `aube install` at the workspace
root and then `aube start` from `packages/app`, asserting it does
**not** print `Auto-installing`.
> 
> <sup>Reviewed by [Cursor Bugbot](https://cursor.com/bugbot) for commit
edf9b04. Bugbot is set up for automated
code reviews on this repo. Configure
[here](https://www.cursor.com/dashboard/bugbot).</sup>
<!-- /CURSOR_SUMMARY -->

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* Skip empty bytes_segment so the phase-2 label doesn't render a
  doubled `·  ·` separator on installs without `dist.unpackedSize`
  (Greptile P1).
* Switch ETA to fetch-window throughput: snapshot completed at the
  set_phase("fetching") boundary, then divide remaining work by
  `fetch_elapsed_ms / (completed - baseline)`. Falls back to `ETA …`
  while the baseline or fetch window are missing. Eliminates the
  pessimistic "110s ETA after a 3s resolve" case (Greptile P2).
* Reconcile estimated_bytes after filter_graph: track per-dep_path
  unpacked_size during streaming, then re-sum surviving dep_paths
  via the new InstallProgress::reconcile_estimated_bytes API.
  Mirrors the package-denominator reconcile applied next to it
  (Greptile P3).
* Apply TARBALL_COMPRESSION_RATIO in TTY refresh_bytes_segment too —
  CI mode applied it; TTY was showing the raw unpacked sum (~3.3×
  inflated) (Cursor #1).
* Refresh ETA on every inc_downloaded_bytes so it doesn't freeze
  during long downloads when no inc_total / inc_reused events fire.
  Documents the trade-off at FetchRow::drop where a refresh would
  require a larger refactor (Cursor #3).
* Style only the ✓ green on the install / no-op summaries instead
  of the whole sentence; aligns with the CI heartbeat summary's
  styling so the success cue stays sharp without bleeding green
  across the line (Cursor #4 + user request).
* Standardize the ✓ prefix on both Already-up-to-date paths
  (`print_install_summary` and `print_already_up_to_date`).

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>