Cleanup with performance gain

[NickSdot]

Adds AbstractSniff::elementNameRanges() to centralises math and deduplicate. Plus shaves roughly 12% off sniffing and 10% off fixing. Before this, we materialised complete lines and source ranges for every candidate. Now we first only do line numbers; ranges are only created for candidates that will be promoted to actual violations.

Depends on #29, #30, #31, #32, #33, #34, #35, #36, #37, #38 and #39.
This PRs diff will shrink as preceding PRs are merged and this branch is rebased.

Diff against previous branch: NickSdot:stack-08-cleanup...NickSdot:stack-09-performance (different to diff in here only until the previous branch was merged)

Version	command	Files	Violations	Runtime
main	.	11,527	115,113	4.19s
stack-08-cleanup	.	11,527	45,171	4.15s
stack-09-performance	.	11,527	45,171	3.67s
stack-09-performance	. --wide	11,527	45,171	3.64s

Since this wasn't explained before: the 115,113 vs 45,171 difference comes from entity handling:

• main reports violations from entity expansion against every referencing file -> duplicates
• we now exclude copies and report them against their actual source files
• --wide adds those referenced source files as targets; hwoever it does not report duplicates like main