improve string handling performance

[phraktle]

Performance of string handling methods, such as ImGui.text or ImGui.calcTextSize etc could be improved by avoiding copying using JNI GetStringCritical instead of GetStringUTFChars. This copying can be quite an overhead on text-heavy user interfaces.

[SpaiR]

GetStringUTFChars converts Java UTF-16 text to native UTF-8. Although GetStringCritical gives us jchar*. Even that ImGui supports UTF-16: in a form ImWchar type, yet it's not quite what we need for everything, like ImGui#textwhich requires char*.

We could pass raw bytes[] from Java, yet it still needs conversions to be done. Do you have thoughts about that?

[phraktle]

Hm... A further complication is Java 9 compact strings, which encodes LATIN1 strings on 8 bits, the rest as UTF16 – neither of which is UTF8 :) I guess ASCII strings could be used directly – it's probably still more efficient to detect if the string is ASCII only (no high bits) than it is to allocate/copy/release a new buffer.

[phraktle]

Maybe another way to reduce the overhead is to do the copying conversion into a pre-allocated buffer, eg. using GetStringCritical + ImTextStrToUtf8, instead of Get/ReleaseStringUTFChars (which allocates a new one, AFAIK)