Precompiling failed on raspberry pi

```
[ Info: Precompiling PythonCall [6099a3de-0909-46bc-b1f4-468b9a2dfc0d]
CondaPkg Found dependencies: /home/pi/.julia/packages/PythonCall/B5irB/CondaPkg.toml
ERROR: InitError: MicroMamba does not support your platform
during initialization of module C
```
I get this error during precompilation. I can do `using PythonCall` no problem, but get this error when trying to import something:
```
julia> pyimport("numpy")
ERROR: UndefRefError: access to undefined reference
Stacktrace:
[1] PyUnicode_DecodeUTF8 at /home/pi/.julia/packages/PythonCall/B5irB/src/cpython/pointers.jl:290 [inlined]
[2] pystr_fromUTF8(::Ptr{UInt8}, ::Int32) at /home/pi/.julia/packages/PythonCall/B5irB/src/concrete/str.jl:1
[3] pystr_fromUTF8 at /home/pi/.julia/packages/PythonCall/B5irB/src/concrete/str.jl:2 [inlined]
[4] pystr at /home/pi/.julia/packages/PythonCall/B5irB/src/concrete/str.jl:10 [inlined]
[5] Py at /home/pi/.julia/packages/PythonCall/B5irB/src/Py.jl:138 [inlined]
[6] macro expansion at /home/pi/.julia/packages/PythonCall/B5irB/src/Py.jl:128 [inlined]
[7] pyimport(::String) at /home/pi/.julia/packages/PythonCall/B5irB/src/concrete/import.jl:11
[8] top-level scope at REPL[15]:1
```

I get that this package is probably not made with the RPi in mind. But would it be possible to support it?

Version info:
```
julia> versioninfo()
Julia Version 1.5.3
Platform Info:
OS: Linux (arm-linux-gnueabihf)
CPU: unknown
WORD_SIZE: 32
LIBM: libopenlibm
LLVM: libLLVM-9.0.1 (ORCJIT, cortex-a72)
Environment:
JULIA_EDITOR = code
JULIA_NUM_THREADS =
JULIA_CONDAPKG_BACKEND = Null
```

Start by reproducing precompilation on the reported Raspberry Pi environment, using the MicroMamba error and the `pyimport("numpy")` failure as entry points. Read the referenced `src/cpython/pointers.jl`, `src/concrete/str.jl`, `src/Py.jl`, and `src/concrete/import.jl` paths, then verify that precompilation completes and importing NumPy no longer raises `UndefRefError`.

Rédigé par le modèle d'indexation à partir du texte de l'issue.