Adding buffer & channel for Marshall & UnMarshall, fixing performance…

[lchanouha]

… issues & memory usage which large files

Hello,
I use your library to import / export LDIF with > 150K entries.

This commit addresses:

• performance issues by replacing basic string concatenation with buffer concat
• io management
• added intermediate function to get entries with a channel, without building a huge ldif.LDIF table.
• error handling, read error is added in the entry, and now skippable

I didn't change the behavior of Marshall and Unmarshall (same spec and error handling), to read / write
a lot of entries, theses functions are prefered:

• func MarshalBuffer(l *LDIF, output io.Writer) (err error) -> to write to an buffer (memory, file, stdout..)
• func UnmarshalBuffer(r io.Reader, l *LDIF) (error, chan *Entry) -> to output entries in a channel.

This is one on my first go code, maybe this commit should me checked (go test passed).

Regards,

[adamluzsi]

I just realised that I did similarly, except using the iterator pattern, to achieve the same streaming idiom within a single goroutine.

#25

[cpuschma]

@adamluzsi Yes, but imo your implementation is cleaner. I'll close this PR once yours is merged. Thank you both anyways for your work! ❤️

[adamluzsi]

The writing / Encoder idiom still valuable for scavenging, but it might worth to prove its behaviour with tests