Bugfix/esp8266 http client

[Jeroen88]

See PR #6176, this PR was merged while there was still an issue to solve.

With this PR:

• ::getString() no more returns an error but returns an empty String() if the server sends a '204 no content'
• if a connection is being reused and this connection gets lost (this can be detected because e.g. ::GET() returns a value less than zero), the connection can be restored by calling ::end() followed by ::begin(). An example of this usage is added as ReuseConnectionV2.ino

This behavior of this example is tested with both a WiFiClient and a (BearSSL) WiFiClientSecure. The lost connection was simulated with switching off the router and switching it on again, and with stopping my HTTP(S) server and restarting it. In the first case a connection refused was reported, in the second case a connection lost. Switching the router back on or restarting the server caused the program to resume showing the ::GET() payload as expected.

[Jeroen88]

@d-a-v as promised a new PR.

Could you take a look at ::disconnect() here and here?
I think the two reset()'s should only be done if !preserveClient and should be added to the second block of code as well. What do you think?

[Jeroen88]

Last commit fixes style check Travis build error. Other Travis failed tests are not caused by my PR

[Jeroen88]

@d-a-v Have you got any idea why the Travis build is failing?

[earlephilhower]

Arduino IDE nightly just changed, breaking our builds. We're looking into it.

[Jeroen88]

@earlephilhower Thnx for your reaction Earle!

[earlephilhower]

mklittlefs.exe is back in the PR, somehow.

[Jeroen88]

@earlephilhower I do not know where it comes from, but I will try to get it out. Could you have a look at this question:

@d-a-v as promised a new PR.

Could you take a look at ::disconnect() here and here?
I think the two reset()'s should only be done if !preserveClient and should be added to the second block of code as well. What do you think?

[Jeroen88]

@earlephilhower please help with git: now I got rid of mklittlefs but got lwip2/builder instead...

[earlephilhower]

You're in git hell, I think. I'd roil back commits to before you did a push containing lwip and try again.

git reset --hard 85d836b # Go back to GH merge a while back
git rm <path to the mklittlefs.exe you've still got here in this hash> # get rid of mkfs
git pull upstream master # merge w/upstream
git submodule update # will pull in the lwip change in the master branch which you missed and remove from the commit later
git commit -a
git push --force

[Jeroen88]

Close because of git problems. Created a new PR #6476