Optimize FromUnderscoreDashSeparatedWords to reduce allocations

[alexzzzs]

Optimize FromUnderscoreDashSeparatedWords to reduce allocations

Optimizes the FromUnderscoreDashSeparatedWords method to eliminate unnecessary memory allocations.

The original implementation (string.Join(" ", input.Split(['_', '-']))) allocates multiple intermediate strings and arrays for each call.

Changes

• Replaced string.Join(" ", input.Split(['_', '-'])) with string.Create for .NET 7.0+ (in-place modification)
• Added Replace fallback for older frameworks
• Zero temporary allocations, O(N) time complexity

Testing

• All unit tests pass
• Builds successful for .NET 8.0, 10.0, and 4.8

[hazzik]

Why not https://learn.microsoft.com/en-us/dotnet/api/system.memoryextensions.replace ?

[alexzzzs]

I wrote the manual loop to handle both _ and - in a single pass to avoid iterating over the span twice.
Although, I didn't realise that MemoryExtensions.Replace uses SIMD/vectorization, which would likely make up for that.
I updated the code to use MemoryExtensions.Replace.

[Copilot]

Pull request overview

This PR optimizes the FromUnderscoreDashSeparatedWords method to reduce memory allocations during string humanization. The optimization is particularly beneficial when processing strings with underscores or dashes.

Key changes:

• Replaced string.Join(" ", input.Split(['_', '-'])) with more efficient implementations
• Added string.Create for .NET 7.0+ to enable zero-allocation in-place character replacement
• Used chained Replace calls as a fallback for older frameworks