Test failures on PowerPC arch

@adamretter alerted us that `make check` is failing on PPC64le (powerpc 64-bit little-endian) architecture on RocksDB master HEAD. On investigating further, I noticed a couple of tests have been failing since RocksDB 5.7 release.
The two specific tests that are failing are:
- FullBloomTest.FullVaryingLengths
- DBPropertiesTest.AggregatedTablePropertiesAtLevel

I narrowed the issue down to be due to ccf5f08f882038e8b481fafa0a0c0b1a04f6a390 introduced in #2579 . The tests work without this commit and fail with this commit included.

### Steps to reproduce the behavior
Either
`make check`
OR
```
make bloom_test db_properties_test
TMP_DIR=/dev/shm ./bloom_test --gtest_filter=FullBloomTest.FullVaryingLengths
TMP_DIR=/dev/shm ./db_properties_test --gtest_filter=DBPropertiesTest.AggregatedTablePropertiesAtLevel
```
Here's the output:

```
ubuntu@hortonworks-2:/tmp/rdb/rocksdb$ make bloom_test db_properties_test -j8
...
...
ubuntu@hortonworks-2:/tmp/rdb/rocksdb$ TMP_DIR=/dev/shm ./bloom_test --gtest_filter=FullBloomTest.FullVaryingLengths
Note: Google Test filter = FullBloomTest.FullVaryingLengths
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from FullBloomTest
[ RUN ] FullBloomTest.FullVaryingLengths
False positives: 0.00% @ length = 1 ; bytes = 133
False positives: 0.00% @ length = 2 ; bytes = 133
False positives: 0.00% @ length = 3 ; bytes = 133
False positives: 0.00% @ length = 4 ; bytes = 133
False positives: 0.00% @ length = 5 ; bytes = 133
False positives: 0.00% @ length = 6 ; bytes = 133
False positives: 0.00% @ length = 7 ; bytes = 133
False positives: 0.00% @ length = 8 ; bytes = 133
False positives: 0.00% @ length = 9 ; bytes = 133
False positives: 0.00% @ length = 10 ; bytes = 133
False positives: 0.00% @ length = 20 ; bytes = 133
False positives: 0.00% @ length = 30 ; bytes = 133
False positives: 0.02% @ length = 40 ; bytes = 133
False positives: 0.04% @ length = 50 ; bytes = 133
False positives: 0.15% @ length = 60 ; bytes = 133
False positives: 0.27% @ length = 70 ; bytes = 133
False positives: 0.59% @ length = 80 ; bytes = 133
False positives: 0.83% @ length = 90 ; bytes = 133
False positives: 1.17% @ length = 100 ; bytes = 133
util/bloom_test.cc:287: Failure
Expected: (FilterSize()) <= ((size_t)((length * 10 / 8) + 128 + 5)), actual: 389 vs 383
200
[ FAILED ] FullBloomTest.FullVaryingLengths (8 ms)
[----------] 1 test from FullBloomTest (8 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test case ran. (8 ms total)
[ PASSED ] 0 tests.
[ FAILED ] 1 test, listed below:
[ FAILED ] FullBloomTest.FullVaryingLengths

1 FAILED TEST

ubuntu@hortonworks-2:/tmp/rdb/rocksdb$ TMP_DIR=/dev/shm ./db_properties_test --gtest_filter=DBPropertiesTest.AggregatedTablePropertiesAtLevel
Note: Google Test filter = DBPropertiesTest.AggregatedTablePropertiesAtLevel
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from DBPropertiesTest
[ RUN ] DBPropertiesTest.AggregatedTablePropertiesAtLevel
db/db_properties_test.cc:200: Failure
Expected: (static_cast(dbl_a - dbl_b) / (dbl_a + dbl_b)) < (bias), actual: 0.599198 vs 0.5
...
```

Arch and OS details on the machine I have access to shows:
```
ubuntu@hortonworks-2:/tmp/rdb/rocksdb$ uname -a
Linux hortonworks-2 4.4.0-101-generic #124-Ubuntu SMP Fri Nov 10 18:29:11 UTC 2017 ppc64le ppc64le ppc64le GNU/Linux
ubuntu@hortonworks-2:/tmp/rdb/rocksdb$
ubuntu@hortonworks-2:/tmp/rdb/rocksdb$ cat /proc/cpuinfo
processor : 0
cpu : POWER8E (raw), altivec supported
clock : 3425.000000MHz
revision : 2.1 (pvr 004b 0201)
...
timebase : 512000000
platform : pSeries
model : IBM pSeries (emulated by qemu)
machine : CHRP IBM pSeries (emulated by qemu)
```