Use lookup table in parse_npy_datatype to avoid complex regex

[Bisaloo]

No description provided.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if f424263 is merged into main:

• ✔️pkg_load: 5.29ms -> 5.26ms [-4.3%, +3.1%]
• 🚀read_bigendian: 11.1ms -> 11.1ms [-1.02%, -0.39%]
• ✔️read_bool: 13.2ms -> 13.2ms [-6.04%, +5.54%]
• ✔️read_empty: 11.2ms -> 11.1ms [-0.72%, +0.07%]
• ✔️read_float32: 12.6ms -> 12.8ms [-1.56%, +3.6%]
• ✔️read_float64: 13ms -> 12.8ms [-5.16%, +1.78%]
• ✔️read_int16: 11.6ms -> 11.5ms [-1.44%, +0.66%]
• ✔️read_int32: 11.7ms -> 11.7ms [-1.54%, +1.53%]
• ✔️read_int64: 12.5ms -> 12.4ms [-1.83%, +0.71%]
• ✔️read_int8: 11.7ms -> 11.5ms [-3.09%, +0.32%]
• 🚀read_npz: 12.9ms -> 12.8ms [-1.14%, -0.37%]
• 🚀read_string: 11.7ms -> 11.5ms [-2.06%, -0.6%]
• 🚀read_structured: 12ms -> 11.9ms [-1.48%, -0.45%]
• ✔️read_uint16: 12.1ms -> 12ms [-1.75%, +0.81%]
• ✔️read_uint32: 11.8ms -> 11.7ms [-1.47%, +0.64%]
• ✔️read_uint64: 11.8ms -> 11.7ms [-1.19%, +0.21%]
• ✔️read_uint8: 11.7ms -> 11.6ms [-1.55%, +0.74%]
• ✔️read_unicode: 11.6ms -> 11.5ms [-1.12%, +0.2%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if c6e656c is merged into main:

• ✔️pkg_load: 5.39ms -> 5.44ms [-0.69%, +2.71%]
• ✔️read_bigendian: 12.1ms -> 12ms [-1.57%, +0.25%]
• ✔️read_bool: 12.7ms -> 12.8ms [-1.11%, +2.35%]
• 🚀read_empty: 11.8ms -> 11.6ms [-2.66%, -0.12%]
• ✔️read_float32: 12.8ms -> 12.8ms [-3.01%, +2.74%]
• ✔️read_float64: 12.8ms -> 12.6ms [-4.07%, +0.5%]
• ✔️read_int16: 12.8ms -> 12.8ms [-1.3%, +0.68%]
• ✔️read_int32: 12.6ms -> 12.8ms [-1.61%, +5.97%]
• ✔️read_int64: 13.2ms -> 13.2ms [-0.7%, +0.56%]
• ✔️read_int8: 12.8ms -> 12.9ms [-1.69%, +4.83%]
• ✔️read_npz: 13.5ms -> 13.4ms [-2.57%, +1.29%]
• 🚀read_string: 12.2ms -> 12ms [-3.93%, -0.42%]
• ✔️read_structured: 13.6ms -> 13.5ms [-1.21%, +0.85%]
• ✔️read_uint16: 11.8ms -> 11.7ms [-1.02%, +0.77%]
• ✔️read_uint32: 12.3ms -> 12.2ms [-4.47%, +3.05%]
• ✔️read_uint64: 12.2ms -> 12.2ms [-1.26%, +0.49%]
• ✔️read_uint8: 11.7ms -> 11.7ms [-1.3%, +0.56%]
• ✔️read_unicode: 12.1ms -> 11.9ms [-2.49%, +0.27%]
  Further explanation regarding interpretation and methodology can be found in the documentation.