Add HPyUnicode_FromFormat(v) and HPyErr_Format.

[steve-s]

Git log message:

 Implement API helper functions  HPyUnicode_FromFormat(V) and HPyErr_Format
    
    The HPy implementation was taken from CPython, but was adapted to be to operate
    only with utf-8 encoding, because that is how HPy exposes Python strings
    (i.e., HPyUnicode_AsUTF8AndSize). CPython starts with UCS1 and widens to
    UCS2,3,4 depending on the largest character encountered so far. We keep
    everything in utf-8, or in fact in ascii as long as we do not encounter any of
    the Python specific formatting units such as %S or %V. The only place where
    we actually need to have special handling for wider characters is with the width
    and precision flags for Python specific formatting units.
    
    The contract is bit stricter than CPython, for example: width/precision are not
    just ignored for %c and %p, but cause an error. NULL and HPy_NULL values are
    not just asserted, but cause a Python level system error. From quick check of
    the top 4000 packages it seems that this should not cause any compatibility
    issues.
    
    We chose to not expose HPyErr_FormatV yet, because it does not seem necessary:
    no usages in top 4000 packages.

[mattip]

I wonder what this means for other implementations. Did you already implement this in GraalPython?

[fangerer]

I wonder what this means for other implementations. Did you already implement this in GraalPython?

There is nothing to implement on GraalPy or on PyPy. The formatting code will be in the user's extension (like HPyArg_Parse and friends). The used ABI (e.g. HPy_Str, etc.) already exists.

[mattip]

Thanks @steve-s