python-hyperscan: Memory leak in Database object when compiling, dumping and loading.

Thanks for the great library. We discovered a memory leak when a Database object is compiled, dumped, and loaded multiple times. This is increasingly worse when the database is growing larger. Even creating a new Database object every time will result in a leak.

Compiling one pattern in the same Database object:

import hyperscan as hs

db = hs.Database(mode=hs.HS_MODE_BLOCK)
for i in range(100000):
    db.compile(expressions=[b'test'], ids=[1], flags=[hs.HS_FLAG_ALLOWEMPTY])

Creating a new Database object every compile, slows the leak down.

import hyperscan as hs

for i in range(100000):
    db = hs.Database(mode=hs.HS_MODE_BLOCK)
    db.compile(expressions=[b'test'], ids=[1], flags=[hs.HS_FLAG_ALLOWEMPTY])

But when the Database object is dumped (and loaded as well) it speeds the memory usage up.

import hyperscan as hs

db = hs.Database(mode=hs.HS_MODE_BLOCK)
for i in range(100000):
    db.compile(expressions=[b'test'], ids=[1], flags=[hs.HS_FLAG_ALLOWEMPTY])
    b = hs.dumpb(db)

I tried to dig into the C code to find the problem, but I do not have enough expertise to solve the issue. Can someone help or point me in the right direction. Thanks!