CI broken at flutter precache step

[gnprice]

CI is failing, starting earlier today (example), with the following:

Run flutter precache --universal
  flutter precache --universal
  shell: /usr/bin/bash -e {0}
  env:
    JAVA_HOME: /opt/hostedtoolcache/Java_Temurin-Hotspot_jdk/17.0.13-11/x64
    JAVA_HOME_17_X64: /opt/hostedtoolcache/Java_Temurin-Hotspot_jdk/17.0.13-11/x64
fatal: Not a valid object name upstream/master

This appears to be a regression introduced as part of the merge of the engine repo into the main/framework repo upstream:

• Flutter master channel is broken on flutter/codelabs CI flutter/flutter#160558

by an assumption that the latest Flutter version can be found at upstream/master. We clone the repo as a remote named origin (the Git default) and with the branch main, which in the Flutter repo always points to the same commit as master.

[gnprice]

This has reappeared with a slightly different symptom (example):

Run flutter precache --universal
  flutter precache --universal
  shell: /usr/bin/bash -e {0}
  env:
    JAVA_HOME: /opt/hostedtoolcache/Java_Temurin-Hotspot_jdk/17.0.13-11/x64
    JAVA_HOME_17_X64: /opt/hostedtoolcache/Java_Temurin-Hotspot_jdk/17.0.13-11/x64
fatal: Not a valid object name origin/master

The upstream script now assumes it can find the latest Flutter version at either upstream/master or origin/master. We still need a version of the workaround in #1176, because we clone the repo with the branch main, which in the Flutter repo always points to the same commit as master.

The reason the issue reappeared is that the upstream PR addressing this last night (flutter/flutter#160574) defeated our original hacky workaround in #1176: it looks for an actual remote named upstream, and when that's not found it looks at origin/master.