[shadow] safe_mode bypassed by links

_This is a _shadow issue* for Issue 51 on Google Code (from which this project was moved).
Added 2011-01-19T00:10:59.000Z by he...@precheur.org.
Labels: Type-Defect, Priority-Medium.
Please make updates to the bug there.*

Original description

>>> from markdown2 import Markdown
>>> m = Markdown(safe_mode='replace')
>>> m.convert("[evil][evil]\n[evil]: javascript:alert('evil');")
u'<p><a href=&quot;javascript:alert(\'evil\');&quot;>evil</a></p>\n'
>>> m.convert('![img](javascript:void\(0\);)')
u'<p><img src=&quot;javascript:void(0);&quot; alt=&quot;img&quot; /></p>\n'

As you can see, it's possible to execute javascript code via the the src & href attributes and maybe others. It would be good to have something like `is_safe_link` in Markdown's contructor:

  def is_safe_link(value):
    return value.startswith('http://') or value.startswith('#')

<b>What version of the product are you using? On what operating system?</b>

1.0.1.17

Start with the original Google Code issue linked in the shadow report and reproduce the two Markdown.convert examples using Markdown(safe_mode='replace'). Review the safe_mode handling and the proposed is_safe_link entry point; done means javascript URLs in link and image attributes are no longer accepted or emitted when safe mode is enabled.