xkbcomp/scanner: avoid extra copies for keynames, keywords, identifiers

[bluetech]

The tokens don't have escapes so no need to use the buf for them.

[wismill]

I run the bench but the improvement is a bit disappointing:

Runtime:

# Master
$ ./build_master/bench-compile-keymap --stdev 0.3 --layout us,de --variant ,neo
mean: 2094 µs; stdev: 0.296366% (target: 0.300000%); last run: compiled 2048 keymaps in 4.292202s; total time: 8.582081s
# This PR
$ ./build_release/bench-compile-keymap --stdev 0.3 --layout us,de --variant ,neo
mean: 2098 µs; stdev: 0.299656% (target: 0.300000%); last run: compiled 1024 keymaps in 2.149264s; total time: 4.299814s

Memory:

$ memusage ./build_master/compile-keymap --layout us,de --variant ,neo --test 2>&1 | head -n 7

Memory usage summary: heap total: 977972, heap peak: 300995, stack peak: 18720
         total calls   total memory   failed calls
 malloc|      23921         622500              0
realloc|       2287         295776              0  (nomove:81, dec:0, free:0)
 calloc|        577          59696              0
   free|      28676         977972

# This PR
$ memusage ./build_release/compile-keymap --layout us,de --variant ,neo --test 2>&1 | head -n 7

Memory usage summary: heap total: 977968, heap peak: 300991, stack peak: 18720
         total calls   total memory   failed calls
 malloc|      23921         622496              0
realloc|       2287         295776              0  (nomove:81, dec:0, free:0)
 calloc|        577          59696              0
   free|      28676         977968

Did I miss something?

[bluetech]

Did I miss something?

No, this PR is not much of an optimization. I will post a PR shortly that builds on this one and shows an improvement.

[wismill]

That makes more sense, thanks for the explanation.

Just 2 nitpicks left for me.