fix: wrong data from Blob/Clob when mark/reset is used

[sfeigl]

Hi,

I spotted an incorrect behaviour using LargeObject BlobInputStream mark() and reset() methods.

Both methods do not take the internal buffer in account.

• mark() actually marks the position after the last byte in the buffer
• reset() changes the large object position, but does not clear the buffer. Buffered bytes from the old position are still returned after reset() until the buffer is depleted.

Here is a suggested fix. The reset() method could be more sophisticated by checking wheter parts of the buffer could be reused.

[codecov-io]

Codecov Report

Merging #971 into master will increase coverage by 0.03%.
The diff coverage is 71.42%.

@@             Coverage Diff              @@
##             master     #971      +/-   ##
============================================
+ Coverage     67.38%   67.42%   +0.03%     
- Complexity     3687     3689       +2     
============================================
  Files           170      170              
  Lines         15638    15640       +2     
  Branches       2531     2532       +1     
============================================
+ Hits          10538    10545       +7     
+ Misses         3917     3911       -6     
- Partials       1183     1184       +1