Fix browser freezing when PostgreSQL decides not to highlight any matches

Closes #386 (closed) .

Frontend entered an infinite loop in MarkdownText.addMarks when PostgreSQL decides to return search results from ts_query but doesn't highlight them when requesting ts_headline with the same query. This seems to happen on staging/prod databases and could not be replicated locally even when all data was copied, which leads me to believe it's a locale issue (my laptop is en_US for reference).

Example backend -> frontend response when searching for "bg" (see the "highlighted_words" field)

On staging/prod:

    {
      "text": "I'm not sure if 1 is correct. If we look at the linearity property of Fourier: For any complex numbers a and b, if h(x) = af(x) + bg(x), then \u0125(\u03be) = a \u00b7 f\u0302(\u03be) + b \u00b7 \u011d(\u03be). So it should be 2 \u00b7 F(u,v) and not 2 \u2217 F(u,v).\r\nAnd if we look at 2 as a constant function, then the DFT of c(x) = 2 isn't 2 but the dirac-delta impulse at 2. So this still doesn't make sense.",
      "long_id": "85iheehpj8j2ax5g",
      "rank": 0.06079271,
      "author_username": "barlasa",
      "author_displayname": "Ayseg\u00fcl Barlas",
      "highlighted_words": [
        "bg"
      ],
      "answer_long_id": "5c1d583729082200e37306e3",
      "exam_displayname": "FS10",
      "filename": "bdubehxd.pdf",
      "category_displayname": "Visual Computing",
      "category_slug": "VisualComputing",
      "type": "comment"
    },
    {
      "text": "sch\u00e9in Schr\u00ebft\u00a0bg. ",
      "long_id": "m9sxb0sowgsm9tgj",
      "rank": 0.06079271,
      "author_username": "caoral",
      "author_displayname": "Cagri Oral",
      "highlighted_words": [],
      "answer_long_id": "wtno55y3wkl0bh4l",
      "exam_displayname": "FS21",
      "filename": "4pft8mxl.pdf",
      "category_displayname": "Lineare Algebra",
      "category_slug": "LineareAlgebra",
      "type": "comment"
    }

On my local machine:

        {
            "text": "I'm not sure if 1 is correct. If we look at the linearity property of Fourier: For any complex numbers a and b, if h(x) = af(x) + bg(x), then \u0125(\u03be) = a \u00b7 f\u0302(\u03be) + b \u00b7 \u011d(\u03be). So it should be 2 \u00b7 F(u,v) and not 2 \u2217 F(u,v).\r\nAnd if we look at 2 as a constant function, then the DFT of c(x) = 2 isn't 2 but the dirac-delta impulse at 2. So this still doesn't make sense.",
            "long_id": "85iheehpj8j2ax5g",
            "rank": 0.06079271,
            "author_username": "barlasa",
            "author_displayname": "Ayseg\u00fcl Barlas",
            "highlighted_words": [
                "bg"
            ],
            "answer_long_id": "5c1d583729082200e37306e3",
            "exam_displayname": "FS10",
            "filename": "bdubehxd.pdf",
            "category_displayname": "Visual Computing",
            "category_slug": "VisualComputing",
            "type": "comment"
        },
        {
            "text": "sch\u00e9in Schr\u00ebft\u00a0bg. ",
            "long_id": "m9sxb0sowgsm9tgj",
            "rank": 0.06079271,
            "author_username": "caoral",
            "author_displayname": "Cagri Oral",
            "highlighted_words": [
                "bg"
            ],
            "answer_long_id": "wtno55y3wkl0bh4l",
            "exam_displayname": "FS21",
            "filename": "4pft8mxl.pdf",
            "category_displayname": "Lineare Algebra",
            "category_slug": "LineareAlgebra",
            "type": "comment"
        }

answer/comment.highlighted_words from the backend is passed as addMarks(obj, new RegExp(highlighted_words.join("|"))). But when the array is empty, the RegExp is instantiated as new RegExp(""). This instance matches all inputs, even the empty string. So string.match(regex) returns a match with 0 length, causing the while loop within addMarks to loop forever.

This never happened on the old search page (and only happens on QuickSearch) since the old search page uses a wrapper component HighlightedMarkdown that checks for this exact case....

We fix the problem and prevent it in the future by:

• Inside addMarks(), if anyone accidentally does the same mistake in the future, check if the RegExp is empty and return early.
• Merge the functionality of HighlightedMarkdown (only used in 1 other place) into MarkdownText, which prevents developers from passing raw RegEx. They must now pass an array of words to highlight, which can now safely by empty or undefined.