bundle2: avoid unbound read when seeking

Currently, seekableunbundlepart.seek() will perform a read() during
seek operations. This will allocate a buffer to hold the raw data
over the seek distance. This can lead to very large allocations
and cause performance to suffer.

We change the code to perform read(32768) in a loop to avoid
potentially large allocations.

`hg perfbundleread` on an uncompressed Firefox bundle reveals
a performance impact:

! bundle2 iterparts()
! wall 2.992605 comb 2.990000 user 2.260000 sys 0.730000 (best of 4)
! bundle2 iterparts() seekable
! wall 3.863810 comb 3.860000 user 3.000000 sys 0.860000 (best of 3)
! bundle2 part seek()
! wall 6.213387 comb 6.200000 user 3.350000 sys 2.850000 (best of 3)
! wall 3.820347 comb 3.810000 user 2.980000 sys 0.830000 (best of 3)

Since seekable bundle parts are (only) used by bundlerepo, this /may/
speed up initial loading of bundle-based repos. But any improvement
will likely only be noticed on very large bundles.

Differential Revision: https://phab.mercurial-scm.org/D1394

mercurial/bundle2.py

@@ -1415,13 +1415,20 b' class seekableunbundlepart(unbundlepart)'
             newpos = self._pos + offset
         elif whence == os.SEEK_END:
             if not self.consumed:
-                self.read()
+                # Can't use self.consume() here because it advances self._pos.
+                chunk = self.read(32768)
+                while chunk:
+                    chunk = self.read(32768)
             newpos = self._chunkindex[-1][0] - offset
         else:
             raise ValueError('Unknown whence value: %r' % (whence,))
 
         if newpos > self._chunkindex[-1][0] and not self.consumed:
-            self.read()
+            # Can't use self.consume() here because it advances self._pos.
+            chunk = self.read(32768)
+            while chunk:
+                chunk = self.read(32668)
+
         if not 0 <= newpos <= self._chunkindex[-1][0]:
             raise ValueError('Offset out of range')