Avoid duplicate keyword search

[fancycode]

Fixes #1057

The example file contains a huge object with this content:

210891 0 obj
<</Limits [(node00000002) (node00275919)]
/Names [(node00000002) 8591 0 R (node00000006) 190838 0 R (node00000007) 66313 0 R (node00000008) 24685 0 R (node00000009) 208328 0 R (node00000010) 163561 0 R (node00000011) 27134 0 R (node00000012) 126131 0 R (node00000013) 9895 0 R (node00000014) 101713 0 R (node00000015) 64243 0 R (node00000016) 118039 0 R (node00000017)
... continues for more than 5 MBytes ...
(node00275919) 209354 0 R]>>
endobj

This blows up the parsing code in

pdfcpu/pkg/pdfcpu/model/parse.go

Line 1243 in 4e0e0db

func DetectKeywords(line string) (endInd int, streamInd int, err error) {

Here the keywords are searched lots of times after the start is advanced for each string literal inside the array (e.g. (node00000002), etc.).

The PR fixes this by searching only once for the keywords and then adjusting the position if strings / comments are found before them. In my tests the example file now parses in about 3 seconds.

[hhrutter]

Sounds very cool..
Let me double check your patch of this critical code section.

[hhrutter]

Why don't we leave the original method name DetectKeywords and pass in as 2nd parm the context from the caller. does that make any sense?

[hhrutter]

There is a(n old) bug in posFloor:

if pos1 < pos2 {
    return pos2
}

should be

if pos1 < pos2 {
    return pos1
}

[hhrutter]

Happy to merge if you can take care of these 2 issues.

[tmm1]

@fancycode ping