Add support for partial matching to pcre2_substitute

[NWilson]

Fixes #846

[rrthomas]

Thanks very much for this. It looks good on a read-through. The restrictions make sense. (I guess it would be possible if desired to reroute the match & substitution through code that works on the entire input if I wanted to support the use of $_ and $', having detected that condition with the ERROR_PARTIALSUBS error.)

I'll see if I can build my code against this branch now.

[rrthomas]

The only thing I would question so far is the wording: relative to the current state, it's correct to say "Add support for partial matching", but relative to the status quo ante 10.47, it's more "Restore support for partial matching, now properly documented and implemented": after all, it worked fine before, though certain uses would have silently failed.

And I don't suppose I was the only person previously relying on this behavior!

[rrthomas]

Yup, works fine for me. A thousand thanks!

[NWilson]

Thank you for those comments, and confirming that it works.

Also, I do acknowledge that I did break your flow in 10.47, which had been previously working. I apologise again. (It's possible however that no-one else had discovered this flow, due to the fact that it required PCRE2_SUBSTITUTE_MATCHED to allow partial matching.)

[rrthomas]

It's less that I feel ignored (I don't!) and more that it might surprise future pedantic readers of the changelog (like me!) who end up confused and side-tracked by the idea that this functionality is brand new in 10.48.

I put "future me" firmly in this class of people, BTW, once I've forgotten this whole issue about 5 minutes after it gets fixed.