to!SomeString should use formatValue

[9rnsr]

Pull #126 was now merged, so we can remove code duplication between std.conv.to and std.format.formatValue.

[andralex]

Oh okay. Can you think of another case where performance may suffer? We don't have good benchmarking for the time being, and I wouldn't want to pull in something that may slow things down.

[9rnsr]

I have reverted only removing dup/idup optimization, see 8b3f98b.
As far as I see, only it was the lacking from my performance perspective.

---

Additionally, it seems to me that the to!String family is convenient function set, not intended to high performance.
The almost of them requires heap allocation to return new string, so that is essentially equals to using appender.
And, maybe, using appender is slower than array concatenation, but I think it is druntime issue, not Phobos issue.
(I am aware of the current Appender implementation is workaround.)

Finally, indeed, this change will make to!String a bit slower, by a little overhead like FormatSpec, but it seems to me that is less important than removing code duplication.

[andralex]

This looks okay, but we need better deprecation messages than "schedule for deprecation". Kenji, could you please look at how deprecation is handled in other modules and do the same here? Thanks!

[9rnsr]

Rebased and improved soft deprecation message (2.056 release == November 2011 + 6 months).