Hypothetical scenario: Suppose you'd want to get all reads with flag 83=paired reads|first in pair|read reverse mapped| proper pair reads with mapping quality > 30 from a bam file and return output to a bam file:

samtools view -h input.bam | awk 'substr($0,1,1) == "@" || $2 == 83 || $5 >= 30 {print}' | samtools view -bS - > output.bam

How it works?:

samtools view -h input.bam - decodes bam and prints sam file with header to screen
awk 'substr($0,1,1) == "@" - keep all header lines
|| $2 == 83 || $5 >= 30 {print}        - the actual condition you would want to check. for unmapped reads you could use GAWK with bitwise operation and($2, 4) == 4 || and($2, 16) == 16, or something like that.
samtools view -bS          - pipe input S=sam to output b=bam and the "-" mediates samtools to take input from the screen
> output.bam               - simple unix redirection to output file

Basically, you can use the awk line to filter anything you want.

You can use BamTools with its filter utility. You need to write a .json file like this one:

{
"isMapped" : "true",
"mapQuality" : ">50",
"isPaired" : "true"
}

And run command

bamtools filter -in in.bam -out out.bam -script filter.json

See a tutorial here.

I don't see a way to extract TLEN with it, though.