DLSS 5 Autopilot 2.0.2

Fixes only: what the reports on 2.0.1 turned up.

Updating from 2.0.x with update now downloads this build, checks its
SHA-256 against the SHA256SUMS.txt published with the release, and offers
a one-click restart. Your settings, library, hidden games and profiles carry
over.

Fixed

• "Neural rendering did not start" about a game where it was running
  (#311). wilsjo2's OptiScaler fork 0.8.4 prints the model's timing from a
  different function (State::EndGpuTiming DLSS-NR elapsed: ...) and counts
  finished pictures (DLSS-NR finished picture: 3600 frames). The diagnosis
  asked for the word Dispatch beside the duration, so neither line counted
  and an earlier line in the log decided the answer. Any DLSS-NR line with a
  duration, or a count of finished pictures, is the model running. The same
  build's one-frame skips ("did not run: ... this frame"), its
  "NR diagnostic" notes and its "falling back to a blocky enlarge" line are
  not read as the model failing. The work-area
  measurement on the optiscaler route reads elapsed: as well as cost:.
• An install that stopped half way in a folder reached through a link
  (#306). A Steam library moved to another place and linked back (mklink)
  has two spellings; the path check followed the link, the installer asked
  for the file's place under the folder as it was opened, and the install
  ended after the add-on and its shaders, with no ReShade.ini and no model.
  The same fault sat in the backup of a Remix mod's files.
• The window raised an error after typing in a box and changing page
  (#296, #300). Text in the library's search box or the video page's link box,
  then another page and back: the page asked the box that had gone with the
  old page. The text is kept.
• "Diagnosis: Working." in a report about a crash (#294, #295). When a
  watched game closed with a Windows fault record, the window and the toast
  said it crashed and the bug report still carried "Working." The report says
  what the window says.
• A game closed on purpose was called a crash (#289). On the optiscaler
  route, when OptiScaler.log ends with the game handing back its NGX
  resources, or with OptiScaler unloading, and Windows recorded a fault within
  seconds of that, the fault is read as one from closing. It is shown, and it
  does not count against the session or the shared result.
• "When it last ran, the process did not have ..." about a launch that was
  over (#287). The list of what a running game had loaded was read out
  under the last launch even when the game had been started again after the
  list was taken. When ReShade.log shows the game started again more than
  two minutes after the list was taken, the diagnosis leaves the list out;
  the report still prints it with its time.
• "The add-on's own output window could not be created" without the add-on
  saying so (#309). The rule looked for "native presentation" and "failed"
  anywhere in the log; a session that works logs "native presentation
  active", and any "failed" beside it completed the match. It reads the
  add-on's own failure line.
• A long message cut off in a small window (#302). The card was limited
  to the window's height. It grows wider first, the screen is its limit, and
  it stays on the screen when the window moves.
• DLSS5-Feeder 1.16.0-beta.5 changed the helper's line for a ReShade.log it
  could not open (ReShade.log could not be opened, before
  ReShade.log is unavailable); the 32-bit helper's diagnosis reads both.