[wleeaf]

NewReaderEncrypted accepts only %PDF-1.0 to %PDF-1.7 directly followed by CR or LF, so it rejects two kinds of files with "not a PDF file: invalid header":

• Every PDF 2.0 file. Their header is %PDF-2.0 (ISO 32000-2, 7.5.2), and pdfTeX and LuaTeX write it, for example.
• Files with white space between the version and the end of the line. tiff2pdf writes %PDF-1.1 \n (Handle space after header #22), and some writers put a NUL byte there.

The check now accepts %PDF-1.0 to %PDF-1.7 and %PDF-2.0, followed by any PDF white-space character. Anything else is still rejected.

Test plan:

• Added TestNewReaderAcceptsHeaders. It opens a one-page file with each of several headers and checks NumPage, and checks that %PDF-1.8, %PDF-2.1, %PDF-1. and %PDF-1.45 are still rejected. On master it fails for %PDF-2.0\n, %PDF-2.0\r, %PDF-1.1 \n and %PDF-1.2\x00\n.
• go test ./... and go build ./... pass.
• I ran 1,637 PDFs from the qpdf and pdf.js test suites through the reader. 23 files that were rejected for their header now open; 17 of them are PDF 2.0, and qpdf reads 22 without a warning.
  • Their objects match qpdf --json, apart from one odd-length hex string that the reader fails on for a separate reason.
  • NumPage agrees with qpdf on all 23, and GetPlainText returns no error.

Fixes #22.