fix: pick the package name prefix per mutex, not per distro

[Hofer-Julian]

Rolling publishes ros2-<pkg> for mutex 0.19 and published ros-rolling-<pkg>
for everything before it. Since #166 the page picked one prefix per distro, so
on the older mutexes it looked up names that do not exist there and reported 0%
of the index available. Reported by @traversaro in #166.

• Key packages by their bare name, so ros2-rclcpp and ros-rolling-rclcpp
  share a row and each mutex slot comes from whichever spelling that generation
  was built under
• Record the prefix per mutex in the dataset and let the table put it in front
  of every name it shows, so rows read ros-rolling-rclcpp on 0.16 to 0.18 and
  ros2-rclcpp on 0.19, links included
• Drop Rolling's ros-rolling-<pkg> compatibility packages: they only pin the
  ros2-<pkg> of the same version and constrain no mutex, so counting them
  would advertise today's version on every older mutex too
• Match either spelling in the search box

Screenshots

Rolling on ros2-distro-mutex 0.18.0 before, nothing available and every name
spelled ros2-:

The same mutex after, back at 42% with the ros-rolling- names it was built
under:

How Has This Been Tested?

• pixi run update-tables: rolling goes from 762 packages on 0.19 and nothing
  on any older mutex to 758, 756 and 755 on 0.18, 0.17 and 0.16, which is ~42%
  of the index instead of 0%
• noetic, humble, jazzy, kilted and lyrical regenerate to build slots identical
  to what is deployed today
• pixi run lint and pixi run build-docs

[traversaro]

Thanks! We are we tracking from which mutex a distro published names under the new scheme?

[Hofer-Julian]

Thanks! We are we tracking from which mutex a distro published names under the new scheme?

Yes.

So for each mutex you check out the corresponding packages.
If for a package both ros2- and ros-rolling- exist, only ros2- is considered.
Then it checks whether more ros2- or ros-rolling- packages are there for one mutex. The highest number wins.

[traversaro]

Thanks! We are we tracking from which mutex a distro published names under the new scheme?

Yes.

So for each mutex you check out the corresponding packages. If for a package both ros2- and ros-rolling- exist, only ros2- is considered. Then it checks whether more ros2- or ros-rolling- packages are there for one mutex. The highest number wins.

Ah cool, so that works out of the box.