Describe the bug
When using precompile NumPy wheels (available since version 1.17.3) briefcase run fails on linux. The problem is with the dynamic libraries:

Deploying dependencies for ELF file /app/appimage/Hello World/Hello World.AppDir/usr/app_packages/numpy/.libs/libopenblasp-r0-34a18dc3.3.7.so 
ERROR: Could not find dependency: libgfortran-ed201abd.so.3.0.0 
WARNING: Failed to deploy dependencies for ELF file /app/appimage/Hello World/Hello World.AppDir/usr/app_packages/numpy/.libs/libopenblasp-r0-34a18dc3.3.7.so, skipping 

I checked that the libraries are present when installing locally, but ldd does report some weirdness:

lib/python3.8/site-packages/numpy/.libs$ ldd libopenblasp-r0-34a18dc3.3.7.so 
	linux-vdso.so.1 (0x00007ffdc9f35000)
	libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f339369c000)
	libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f3393679000)
	libgfortran-ed201abd.so.3.0.0 => not found
	libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f3393487000)
	/lib64/ld-linux-x86-64.so.2 (0x00007f339558f000)

I checked, the libgfortran lib is present. I'm not sure how NumPy tells the system to look for the library in the local folder, but if I remove the file I can't locally import numpy anymore. Maybe briefcase (via linuxdeploy) can ignore the 'missing dependency'?

To Reproduce
Steps to reproduce the behavior:

1. Create new briefcase project (briefcase new)
2. Add numpy==1.17.3 in the requires field.
3. briefcase create on linux
4. briefcase run on linux
5. See error.

Expected behavior
I expect briefcase to package the app and briefcase run to just complete normally.

Environment:

• Operating System: Ubuntu 20.04
• Python version: 3.8
• Software versions:
  • Briefcase: 0.3.5

Additional context
I suspect the problem to lie with the way linuxdeploy builds the appimage. When I have the time, I'll try to figure out how to use linuxdeploy or appimage and try to package NumPy without briefcase. Maybe the problem lies not with briefcase itself.