Add new exceptions for invalid inputs and deprecate GMTInvalidInput exception

[seisman]

Currently, we have several exceptions defined in https://github.com/GenericMappingTools/pygmt/blob/main/pygmt/exceptions.py.

Below is a simple command to count how often each exception is raised in the codebase (The tests directory and the exceptions.py file are excluded).

for e in \
    GMTError \
    GMTOSError \
    GMTCLibError \
    GMTCLibNotFoundError \
    GMTCLibNoSessionError \
    GMTVersionError \
    GMTInvalidInput \
    GMTImageComparisonFailure; 
do
    n=$(grep -R --exclude-dir=tests --exclude="exceptions.py" "raise $e" pygmt | wc -l)
    echo "$e: $n" 
done

the output is:

GMTError: 1
GMTOSError: 1
GMTCLibError: 16
GMTCLibNotFoundError: 1
GMTCLibNoSessionError: 1
GMTVersionError: 1
GMTInvalidInput: 87
GMTImageComparisonFailure: 1

Most exceptions are about the GMT C library, and is only used once or a few times. But the GMTInvalidInput exception is used 87 times and will definitely increase as we wrap more modules.

The GMTInvalidInput exception is general and doesn't tell exactly what's happening. I haven't checked the source codes carefully, but I feel the exceptions can be grouped into a few categories:

1. A required parameter is not given
2. Given parameters have conflicts
3. A parameter is given with an incorrect type
4. A parameter is given with invalid values
5. Maybe more

I think we should have more accurate exceptions, e.g.,

• GMTParameterError: for cases 1 and 2, maybe having more fine-grain exceptions like GMTRequiredParameterError/GMTConflictParameterError
• GMTTypeError: for case 3, similar to the built-in TypeError
• GMTValueError: for case 4, similar to the built-in ValueError.

The exception may even start with PYGMT, not GMT, since these are not directly related to GMT.