Change in type promotion introduced in Numpy 2.0. Fixes to edf.py. (#527)

As discussed in #493, numpy
2.0 introduced changes to type promotion. The change is outlined at:
https://numpy.org/devdocs/numpy_2_0_migration_guide.html#changes-to-numpy-data-type-promotion

Running the tests in `weak_and_warn` raises the following warnings:

```
tests/io/test_convert.py::TestEdfToWfdb::test_edf_uniform
  /Users/tompollard/projects/wfdb-python/wfdb/io/convert/edf.py:409: UserWarning: result dtype changed due to the removal of value-based promotion from NumPy. Changed from int16 to int64.
    temp_all_sigs[i].flatten() - baseline[i]

tests/io/test_convert.py::TestEdfToWfdb::test_edf_non_uniform
  /Users/tompollard/projects/wfdb-python/wfdb/io/convert/edf.py:420: UserWarning: result dtype changed due to the removal of value-based promotion from NumPy. Changed from int16 to int64.
    (temp_sig_data[start_ind:stop_ind] - baseline[i])

tests/io/test_convert.py::TestEdfToWfdb::test_edf_non_uniform
  /Users/tompollard/projects/wfdb-python/wfdb/io/convert/edf.py:414: UserWarning: result dtype changed due to the removal of value-based promotion from NumPy. Changed from int16 to int64.
    sig_data[:, i] = (temp_sig_data - baseline[i]) / adc_gain_all[i]
```

This pull request addresses the issue by setting `temp_all_sigs` and
`temp_sig_data` to `np.int64`.

For reference, `temp_all_sigs` and `temp_sig_data` are initially set as
`np.int64`:


https://github.com/MIT-LCP/wfdb-python/blob/c6d4fd9f05007ebc5f79dd07332f34410d7b8493/wfdb/io/convert/edf.py#L402-L404

`baseline` is set to `int64` here:


https://github.com/MIT-LCP/wfdb-python/blob/c6d4fd9f05007ebc5f79dd07332f34410d7b8493/wfdb/io/convert/edf.py#L353-L355

Author: tompollard

wfdb/io/convert/edf.py

@@ -402,23 +402,27 @@ def read_edf(
     temp_sig_data = np.fromfile(edf_file, dtype=np.int16)
     temp_sig_data = temp_sig_data.reshape((-1, sum(samps_per_block)))
     temp_all_sigs = np.hsplit(temp_sig_data, np.cumsum(samps_per_block)[:-1])
+
     for i in range(n_sig):
         # Check if `samps_per_frame` has all equal values
         if samps_per_frame.count(samps_per_frame[0]) == len(samps_per_frame):
             sig_data[:, i] = (
-                temp_all_sigs[i].flatten() - baseline[i]
+                temp_all_sigs[i].flatten().astype(np.int64) - baseline[i]
             ) / adc_gain_all[i]
         else:
             temp_sig_data = temp_all_sigs[i].flatten()
+
             if samps_per_frame[i] == 1:
-                sig_data[:, i] = (temp_sig_data - baseline[i]) / adc_gain_all[i]
+                sig_data[:, i] = (
+                    temp_sig_data.astype(np.int64) - baseline[i]
+                ) / adc_gain_all[i]
             else:
                 for j in range(sig_len):
                     start_ind = j * samps_per_frame[i]
                     stop_ind = start_ind + samps_per_frame[i]
                     sig_data[j, i] = np.mean(
-                        (temp_sig_data[start_ind:stop_ind] - baseline[i])
-                        / adc_gain_all[i]
+                        temp_sig_data[start_ind:stop_ind].astype(np.int64)
+                        - baseline[i] / adc_gain_all[i]
                     )
 
     # This is the closest I can get to the original implementation