[Types] Add timedelta64 support

Overview

Add support for timedelta64 type to NumSharp for duration arithmetic and time series operations.

Problem

NumSharp lacks timedelta support, needed for datetime arithmetic:

>>> import numpy as np
>>> np.timedelta64(30, 'D')
numpy.timedelta64(30,'D')

>>> np.timedelta64(2, 'h') + np.timedelta64(30, 'm')
numpy.timedelta64(150,'m')

>>> np.datetime64('2024-01-15') + np.timedelta64(30, 'D')
numpy.datetime64('2024-02-14')

>>> np.array([1, 2, 3], dtype='timedelta64[D]')
array([1, 2, 3], dtype='timedelta64[D]')

Use cases:

• Duration arithmetic — Add/subtract time periods
• Time differences — datetime64 - datetime64 = timedelta64
• Time series — Intervals, sampling rates
• Data science — Age calculations, time-to-event

Proposal

Task List

• Create TimeDelta64 struct with unit awareness:

  public readonly struct TimeDelta64
  {
      public readonly long Value;  // Duration in units
      public readonly DateTimeUnit Unit;  // ns, us, ms, s, m, h, D, W, M, Y
  }
  

• Add NPTypeCode.TimeDelta64 to NPTypeCode.cs
• Implement unit conversion logic (shared with datetime64)
• Implement arithmetic:
  • timedelta + timedelta → timedelta
  • timedelta - timedelta → timedelta
  • timedelta * scalar → timedelta
  • timedelta / scalar → timedelta
  • timedelta / timedelta → float
• Implement comparison operators
• Integrate with datetime64:
  • datetime + timedelta → datetime
  • datetime - timedelta → datetime
  • datetime - datetime → timedelta
• Add np.timedelta64() constructor
• Update type promotion tables
• Add tests with NumPy verification

Units (matching NumPy)

Same as datetime64: Y, M, W, D, h, m, s, ms, us, ns

C# Type Mapping

Implementation Effort

MEDIUM — Similar to datetime64, with additional arithmetic operations.

Estimated: ~300 lines for struct and operations.

Related

• Part of NumPy 2.x alignment: #529
• Will benefit from DynamicMethod IL emission: #544
• Blocked by: datetime64 support (#570) should be implemented together

References

• NumPy timedelta64 docs: https://numpy.org/doc/stable/reference/arrays.datetime.html#datetime-and-timedelta-arithmetic