fix: ADB.pidof integration issue with grep and CRLF edge case

[noomorph]

Resolves #924

[tomatobrown]

any eta on getting this merged?

[noomorph]

@tomatobrown, technically, this PR is ready, but two things did not happen:

1. it was not approved by maintainers
2. and honestly, I'm still confused about what's happening in your environment with CRLF line endings from Android Emulator on Mac. It is not supposed to happen at all. I think Rotem is hesitant to approve also because of this concern.

[noomorph]

Dumpsys power might also emit CRLF output, and this means that we are going into a wrong direction.

[noomorph]

Important

I have erased the previous implementation in favor of a simpler one, using adb exec-out instead of adb shell. It feels right and if it proves to be working across all major platforms, I'll be happy to merge it soon.

[noomorph]

Unfortunately, exec-out had issues of its own, with lower API levels, with handling of ps command (the results were different from adb shell on some of Android versions). So, I'm trying now a third approach - to run grep inside adb shell itself and replace CRLF with LF for every stdout string we get.

Still on it.

[noomorph]

Guys, sorry for the delay. I just have finished the other unrelated but highly important task, so I finally can start sorting out the issues one by one. ETA - tomorrow I make a one more bugfix to PR and merge it

[noomorph]

Generally speaking, in this pull request there are already multiple various fixes:

• adb install throws an error if it went wrong
• pidof() method uses internal Android's grep, eliminating by this CRLF issue at all
• getFileSize() (used by artifacts) uses du instead of wc for a broader compatibility with Android devices
• isFileOpen() (used by artifacts) uses lsof | grep which is a more backward-compatible method
• logcat() (used by log artifact) has a fallback for missing -T <time> parameter in older APIs, and | grep pipe instead of --pid=<pid> parameter
• exec() helper method replaces CRLF endings if any to fix adb.devices() method on Windows.
• telnet calls to emulator are logged (sometimes telnet hangs without any message)

TODOs:

• revert win32Implementation changes in detox/src/utils/pipeCommands.js (leftover from previous PR implementations)
• fix logical error for API level cache in ADB class - it assumes you always do install which is not true
• work around lsof issue on some API levels (cannot access /proc/* files)
• test on API levels 19-28 for Mac, Linux and Windows

I did a lot of testing today manually, unfortunately, CI does not catch dozens of environment integration issues, so I still have to work on this PR tomorrow.