You can do this in R with a combination of the seqinR and the aplpack packages:

library(seqinr)
data=read.fasta("file.fasta")
l=getLength(x)

library(aplpack)
stem.leaf(l)

There are a number of options to suit your needs:

stem.leaf(l, unit=100)

Just have a look here for more output options and details:

?stem.leaf

This functions extends far more than the simple stem() function in the graphics package.

My initial idea was to use Biopython for the parsing of the fasta file and matplotlib for the drawing of the stem-leaf plot. I couldn't find a function that draws a text histogram in matplotlibto my amazement, so here is my take:

from Bio import SeqIO
from collections import defaultdict

input_file = r'path\to\file.fasta'
lengths = [len(record.seq) for record in SeqIO.parse(open(input_file), 'fasta')]
d = defaultdict(list)
for length in lengths:
    d[length/10].append(length%10)

for nr in range(min(d.keys()), max(d.keys())+1):
    print "{0:>3} | {1}".format(nr, ' '.join(map(str, sorted(d[nr]))) if nr in d else '')

For this fictitious input file it produces this histogram:

  5 | 3 4 8
  6 | 0 0 6 7 9
  7 | 0
  8 | 6
  9 | 0
 10 | 9
 11 | 2 5 6
 12 | 0 0 0 0 5 8

There is something in the kent source tree called textHistogram. It takes a file with one number per line, which you can get from fasta using awk...

textHistogram <(awk < test.fa '$0 !~ /^>/ {print length($0)}')