BUG: Use 1MB as offset for readNextEndLine (py-pdf#321)

Try to find “%%EOF” in last 1Mb of file.

This fixes the issue with reading Selenium-generated PDF files.

Closes py-pdf#177
Closes py-pdf#442
Closes py-pdf#480

PyPDF2/pdf.py

@@ -1821,12 +1821,12 @@ def read(self, stream):
         absoluteEndFilePos = stream.tell() + 1
         if not stream.tell():
             raise PdfReadError('Cannot read an empty file')
-        last1K = absoluteEndFilePos - 1024  # offset of last 1024 bytes of stream
+        last1M = stream.tell() - 1024 * 1024 + 1 # offset of last MB of stream
         line = b_('')
         while line[:5] != b_("%%EOF"):
-            if stream.tell() < last1K:
+            if stream.tell() < last1M:
                 raise PdfReadError("EOF marker not found")
-            line = self.readNextEndLine(stream, last1K)
+            line = self.readNextEndLine(stream)
             if debug: print("  line:",line)
 
         # find startxref entry - the location of the xref table