Hotfix - Improve the rebuilding of target XLF files

[didier-84]

We rewrote the writeTargetFiles method in base.js to iterate on existing XML trans-units (based on the source XLF file) instead of iterating on the segments from the Translation.io response.

This ensures that all XML trans-units for which we compute the same uniqueIdentifier will be filled with the same target from the Translation.io response.

This fix is put in place after one of our users noticed that some XML trans-units were not filled with the target during a sync.
Indeed, the angular i18n-extractor creates two separate trans-units for sources that "seem" identical but are actually padded with spaces:

<p i18n>Example</p>

<div>
    <p i18n>
      Example
    </p>
</div>

Our package purposely trims any leading or trailing space when processing sources, so this results in only one segment being sent to Translation.io.