Fixes for loading document from stream

[avonwyss]

The PdfDocument constructor which accepts a stream and the FPDF_LoadDocument methods were confusing, since they did not use their "stream" and "count" parameters, but instead the low-level undocumented "fileRead".

I cleaned this up, and added a small improvement when using MemoryStream streams (direcly copy from buffer instead of reading from stream). Successfully tested the functionality with the Test.Console application:

using (var docStream = new MemoryStream()) {
    using (var srcStream = new FileStream("TestDoc.pdf", FileMode.Open))
        srcStream.CopyTo(docStream);
    docStream.Seek(0, SeekOrigin.Begin);
    using (var doc = new PdfDocument(docStream, 0, "password")) {

[bradleypeet]

This is how I do it in my fork of this project. (I have not submitted a pull request.)

	public PdfDocument(Stream stream, int count = 0, string password = null)
            : this(FPDF_FILEREAD.FromStream(stream, count), password)
        {
        }

        internal PdfDocument(FPDF_FILEREAD reader, string password = null)
            : this(PDFium.FPDF_LoadCustomDocument(reader, password))
        {
            // must maintain reference to reader to prevent the callback delegate (passed as an
            // unmanaged function pointer to native pdfium lib) from getting garbage collected.
            streamReader = reader;
        }

[avonwyss]

            // must maintain reference to reader to prevent the callback delegate (passed as an
            // unmanaged function pointer to native pdfium lib) from getting garbage collected.

That's indeed a very good point! I have adjusted my PR accordingly.