Some issues with the `nickbar' module have surfaced that may be related
to another issue involving unexpected connection loss. Thanks to Corwin
for reporting this.

It appears that the teardown logic in `erc-nickbar-disable' was calling
`dframe-close-frame', `speedbar-set-timer', and `speedbar-timer-fn'
repeatedly and unnecessarily, which resulted in redundant scheduling of
the same update timer (there should only be one per speedbar instance).
Other problems involving the variable `speedbar-update-flag' were also
discovered, as well as a possible bug in `dframe-set-timer', which I'll
leave for future people familiar with dframe to deal with if necessary.