Fix unstable formatting around comments in type declarations

[JohnnyMorganz]

Fixes #397 (ref: #430)

[codecov]

Codecov Report

Merging #435 (e996c06) into master (8b3b1dd) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #435   +/-   ##
=======================================
  Coverage   96.42%   96.42%           
=======================================
  Files          14       14           
  Lines        5228     5228           
=======================================
  Hits         5041     5041           
  Misses        187      187           

Impacted Files	Coverage Δ
src/formatters/trivia.rs	86.36% <ø> (ø)
src/formatters/trivia_util.rs	94.68% <ø> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8b3b1dd...e996c06. Read the comment docs.

[JohnnyMorganz]

@matthargett @flakolefluk @ktrz thoughts on the new snapshots?

The newly added snapshots of your tests looks good, but I'm not 100% on the change in the table snapshot (its happening because there is a comment in between the equals token and the table in the declaration)

[matthargett]

innnnnteresting. it would be nice if it didn't force the open brace onto the next line with an indent, but I also think the trade-off here is a good one since the affected regression case is a weird/rare one.

[ktrz]

I think the change in this one snapshot is a good tradeoff for having the stable tool output.

[flakolefluk]

snapshots look good to me! i agree with the previous comments. Let's start with making the output stable. 🎉

[matthargett]

with this branch, I see an interesting regression where indentation is lost column-width set to 90:

exports.separateDisplayNameAndHOCs =
        function(displayName: string | nil, type_: ElementType): (string | nil, Array<string> | nil)
                if displayName == nil then
                        return nil, nil
                end

                local hocDisplayNames: Array<string>? = nil

                if
                        type_ == ElementTypeClass
                        or type_ == ElementTypeForwardRef
                        or type_ == ElementTypeFunction
                        or type_ == ElementTypeMemo
                then
                        -- ROBLOX deviation: use match instead of indexOf
                        if (displayName :: string):match("%(") then
                                -- ROBLOX deviation: use gmatch instead of /[^()]+/g
                                local matches = (displayName :: string):gmatch("[^()]+")
                                local nextMatch = matches()
                                if nextMatch then
                                        displayName = nextMatch
                                        hocDisplayNames = {}
                                        while (nextMatch :: any) ~= nil do
                                                nextMatch = matches()
                                                table.insert(hocDisplayNames :: Array<string>, nextMatch)
                                        end
                                end
                        end
                end

        end

[JohnnyMorganz]

Turns out the above diff is not related to the PR, but is actually a regression from 0.12.5 -> 0.13.0