ASF Duration Calculation Fix, Creation Date Fix

[benrr101]

Description: Another couple bugs I found while porting ASF support in TagLib# to node 😅

The TagLib# implementation of ASF duration wasn't calculating the same duration for the sample.wma file as FFMpeg, foobar2000, VLC, which all said the file was 0:04 (or 4.153s). TagLib# was reporting the file was around 5s. Digging through the ASF specification and reading the sample with a hex editor, I found two issues with the TagLib# implementation:

1. The ASF File Properties Object (page 14 in the linked PDF) stores the play duration before the send duration. TagLib# is reading/writing the object the other way around, which means duration was calculated based on send duration (which in this case is a hundred milliseconds less than the play duration). This also means TagLib# is potentially corrupting ASF files by writing the send/play durations backwards. Fix for this is to swap the order play/send duration are read/written in the FilePropertiesObject class.
2. Preroll duration in the File Properties Object is stored in milliseconds. TagLib# is reading that value as 100ns ticks, so when it is subtracted from the play duration to calculate duration, it's subtracting a far smaller value than it should. Fix for this is to subtract a TimeSpan constructed from the preroll as milliseconds.

The issue described in #255 states that ASF spec says creation date in File Properties Object is 100ns ticks since 1/1/1601 instead of the C# DateTime format's 100ns ticks since 1/1/0000. I don't really know why this is the case, but it is what it is. The fix for this is to add the number of ticks for 1/1/1601 to the creation date as read from the file. ... However, after revisiting the code, I realized creation date is never used and it's impossible to get access to an instance of the class without manually re-reading it from the AsfFile, so this fix is more/less inconsequential. Nevertheless, it's bug, so I fix 😄. Unrelated, I'm planning to expose the HeaderObject in the AsfFile class in my port, and I'm thinking this would be a valuable addition to the TagLib# API.

Testing:

• Switched the ASF format tests from using the StandardFileTests properties test to checking all the properties we know about the sample file. All cases pass!

[decriptor]

@benrr101 Thanks for the fixes.