Fix the bug of loading shared libraries.

[Xianguang-Zhou]

Purposes:
1. Reduce the CPU usage of "qt.py" for Windows.
2. Fix the bug of loading shared libraries.

[cztomczak]

How does the "external_message_pump" option exactly affect performance? How did you test it? What versions of qt did you test with? (pyqt4, pyqt5, pyside, pyside2).

What is the bug with loading shared libraries?

[Xianguang-Zhou]

1. On Windows, if the "external_message_pump" option is not true, the CPU usage is about 10%. If the "external_message_pump" option is true, the CPU usage is about 0%.
   I tested it with PySide2.

2. If a process overrides the "LD_LIBRARY_PATH" environment variable, the subprocesses will not be able to find other shared libraries by the original "LD_LIBRARY_PATH" environment variable.

[cztomczak]

Please be specific. Tested on what website and what OS? What are you doing in app that it hits 10% CP U?

What other shared libraries CEF Python needs to find in OS through LD_LIBRARY_PATH? Is it your app specific libraries? Provide specific information on your issue.

[Xianguang-Zhou]

1. Tested on Windows 10 1809.
   I did almost nothing in the application: qtcef.py

2. The other shared libraries to find through LD_LIBRARY_PATH are libpython libraries.
   For example: libpython3.6m.so.1.0

[cztomczak]

The patch is for qt.py example, Does the issue reproduce with the original example on google website?

[Xianguang-Zhou]

Sorry, I don't know what the original example on google website you said.

[cztomczak]

The one you're sending the PR for.

[Xianguang-Zhou]

I'm terribly sorry.
The original "qt.py" has no CPU usage issues on Windows 7.
The cause of the CPU usage issue may be a bug of Windows 10 1809.
So I plan to delete the commit: 5b6aed9

[Xianguang-Zhou]

I have deleted the commit of enabling the "external_message_pump" option.

[cztomczak]

I will merge it during next release.

FYI: It might be better to get rid of LD_LIBRARY_PATH entirely and link via rpath/$ORIGIN.

[Xianguang-Zhou]

Thank you very much.