Compute LG SourceRange based on content instead of parsed token

[Danieladu]

Fixes #3076

Root cause

Previously, the sourceRange(start and end line position) of a LG template is calculated based on the parsed token. But in certain scenarios that, when there are extra empty character (line breaks) at the end of a template, will cause this calculation not correct, and thus causing the further update on this lg file misplaced in the wrong position, causing potential data loss.

Fix

Use the real content received to calculate start\end line. Content is the golden truth, more trustworthy than parsed tokens.

Impact

Composer

[coveralls]

Pull Request Test Coverage Report for Build 371449104

• 9 of 9 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.02%) to 84.785%

---

Totals
Change from base Build 370698341:	0.02%
Covered Lines:	17312
Relevant Lines:	19498

---

💛 - Coveralls