Speed up "Load into vimscript".

[xavierd]

From 0.012s to 0.004s.

I'm using pyeval() in vim to avoid all the str() call in python. That way, no conversion is done. However, this requires an up to date vim, as noted in #270.

[tobiasgrosser]

I think it would be unfortunate to require everybody to compile their own vim versions. This adds a lot of overhead,
for a very limited benefit. My vim version supports threads very well and works flawless with the current clang_complete. The only missing featue is pyeval. I just tried this patch with a manual installation. It gives
a speedup from 0.007s to 0.004s. Given that I have an overall completion time of 0.138783, this does not make a big difference. From my point it seems we can get along very well without using pyeval. However, if you think the
speedup is still wanted, we can probably support both the old and the new vim at the same time. Your change request has only four changed lines. Making them conditional on the availability of pyeval should not be too involved.