{Feature} Visualization - Auto-show optional panels on first stream

Summary:
Explanation:
`aria_streaming_viewer` sets up its Rerun blueprint once from an
`AriaDataViewerConfig` snapshotted at construction time. GPS and
Neural Band panels stay hidden even when the device is actively
publishing those streams, while the Fixation Crop / Cropped POV
panels are hardcoded on regardless of the recording profile. It
cannot mirror the offline `aria_rerun_viewer` up front because the
Client SDK offers no API to enumerate the sensors a streaming session
will actually push (only the profile name is exposed;
`RecordingProfile` in `sdk_gen2.pyi` carries no sensor bits).

Fix follows the pattern already used in
`multi_device_streaming_viewer`: observe stream first-fires and
rebuild the blueprint via `rr.send_blueprint()`.

- `aria_data_plotter.py` gains
  `AriaDataViewer.set_optional_panel_visible(panel, visible)` — a
  public, idempotent seam that flips one of `enable_gps` /
  `enable_neural_band_batch` / `enable_crop_visualization` and calls
  `update_rerun_blueprint()` when the flag actually changed. Respects
  `blueprint_path` (custom blueprint mode) and defers when the first
  calibration has not yet arrived, in which case the flag is still
  updated so it takes effect on the next blueprint build.

- `aria_streaming_viewer.py` drops the hardcoded
  `enable_crop_visualization = True`, tracks a `_seen_optional_panels`
  set, and calls `set_optional_panel_visible(panel, True)` on the first
  arrival of `gps` / `neural_band_batch` / `fixation_crop` /
  `cropped_pov_image` data. Additive-only: panels can appear during a
  session but do not disappear, mirroring the offline viewer's
  behavior. Rebuild is bounded to one call per optional panel per
  session (idempotency at both layers); callback-thread races at worst
  cause a duplicate rebuild.

Trade-off: optional panels pop in during the first few milliseconds
of their stream instead of being present at spawn. This is
intentional and mirrors the offline viewer, which gates its Neural
Band panel by VRS-content detection at open time — the streaming case
has no equivalent up-front knowledge, so it detects empirically.

Reproducibility:
Streamed a Gen2 profile that publishes GPS and fixation crop but not
Neural Band; before: crop panels visible with no data, GPS panel
absent despite GPS ticking. After: GPS panel appears once the first
GPS sample lands; crop panels appear once the first crop lands; the
Neural Band panel stays absent for the whole session.

Reviewed By: nrraina

Differential Revision: D114112124

fbshipit-source-id: f5826caf90eeaecef02080aee508e94b60bb78c3

Author: kongchen1992

projectaria_tools/tools/aria_rerun_viewer/aria_data_plotter.py

@@ -341,6 +341,36 @@ def set_device_calibration(self, device_calibration):
             self.update_rerun_blueprint()
         self.plot_device_extrinsics()
 
+    OPTIONAL_PANEL_CONFIG_FLAGS = {
+        "gps": "enable_gps",
+        "neural_band_batch": "enable_neural_band_batch",
+        "crop_visualization": "enable_crop_visualization",
+    }
+
+    def set_optional_panel_visible(self, panel: str, visible: bool) -> None:
+        """Show or hide an optional panel, rebuilding the blueprint if it changed.
+
+        Idempotent: repeated calls with the same value are no-ops. When a custom
+        blueprint is in use or the first calibration has not yet arrived, the
+        config flag is still updated so it takes effect on the next blueprint
+        build, but no rebuild is triggered here.
+
+        panel: one of the keys in `OPTIONAL_PANEL_CONFIG_FLAGS`
+            (`"gps"`, `"neural_band_batch"`, `"crop_visualization"`).
+        """
+        if panel not in self.OPTIONAL_PANEL_CONFIG_FLAGS:
+            raise ValueError(
+                f"Unknown optional panel {panel!r}; expected one of "
+                f"{sorted(self.OPTIONAL_PANEL_CONFIG_FLAGS)}"
+            )
+        flag = self.OPTIONAL_PANEL_CONFIG_FLAGS[panel]
+        if getattr(self.config, flag) == visible:
+            return
+        setattr(self.config, flag, visible)
+        if self.config.blueprint_path or self.device_calibration is None:
+            return
+        self.update_rerun_blueprint()
+
     def _get_plot_color(self, plot_label):
         """Helper function to get the color for 2D plots."""
         if plot_label in self.PLOT_COLORS_AND_SIZES_2D: