Pin `numpy` to 1.X.X

[kennyworkman]

Downloading pyensembl in a fresh environment yielded:

Traceback (most recent call last):  File "/opt/mamba/envs/pyensembl/bin/pyensembl", line 8, in <module>
    sys.exit(run())
  File "/opt/mamba/envs/pyensembl/lib/python3.10/site-packages/pyensembl/shell.py", line 281, in run
    genome.index(overwrite=args.overwrite)
  File "/opt/mamba/envs/pyensembl/lib/python3.10/site-packages/pyensembl/genome.py", line 280, in index
    self.db.connect_or_create(overwrite=overwrite)
  File "/opt/mamba/envs/pyensembl/lib/python3.10/site-packages/pyensembl/database.py", line 286, in connect_or_create
    return self.create(overwrite=overwrite)
  File "/opt/mamba/envs/pyensembl/lib/python3.10/site-packages/pyensembl/database.py", line 206, in create
    df = self._load_gtf_as_dataframe(
  File "/opt/mamba/envs/pyensembl/lib/python3.10/site-packages/pyensembl/database.py", line 613, in _load_gtf_as_dataframe
    df = read_gtf(
  File "/opt/mamba/envs/pyensembl/lib/python3.10/site-packages/gtfparse/read_gtf.py", line 257, in read_gtf
    result_df = result_df.to_pandas()
  File "/opt/mamba/envs/pyensembl/lib/python3.10/site-packages/polars/dataframe/frame.py", line 2093, in to_pandas
    return self._to_pandas_without_object_columns(
  File "/opt/mamba/envs/pyensembl/lib/python3.10/site-packages/polars/dataframe/frame.py", line 2144, in _to_pandas_without_object_columns
    record_batches = df._df.to_pandas()
  File "/opt/mamba/envs/pyensembl/lib/python3.10/site-packages/pyarrow/__init__.py", line 65, in <module>
    import pyarrow.lib as _lib
AttributeError: _ARRAY_API not found

It seems like pyarrow is not pinning numpy correctly, but I think you'll have a hard time making changes in that library in a timely fashion.
Would enforce in gtfparse. Reinstalling to the latest 1.X.X, eg:

pip install numpy==1.24.3

Fixes the issue.